Cognition
seventh edition
Cognition
exploring the science of the mind
Daniel Reisberg
reed college
n
W. W. Norton & Company
New York • London
7e
W. W. Norton & Company has been independent since its founding in 1923, when William Warder
Norton and Mary D. Herter Norton first published lectures delivered at the People’s Institute,
the adult education division of New York City’s Cooper Union. The firm soon expanded its program beyond the Institute, publishing books by celebrated academics from America and abroad.
By midcentury, the two major pillars of Norton’s publishing program—trade books and college
texts—were firmly established. In the 1950s, the Norton family transferred control of the company to its employees, and today—with a staff of four hundred and a comparable number of
trade, college, and professional titles published each year—W. W. Norton & Company stands as
the largest and oldest publishing house owned wholly by its employees.
Copyright © 2019, 2016, 2013, 2010, 2006, 2001, 1997 by W. W. Norton & Company, Inc.
All rights reserved
Printed in the United States of America
Editor: Ken Barton
Project Editor: David Bradley
Assistant Editor: Eve Sanoussi
Editorial Assistant: Katie Pak
Manuscript Editor: Alice Vigliani
Managing Editor, College: Marian Johnson
Managing Editor, College Digital Media: Kim Yi
Production Managers: Ashley Horna and Jane Searle
Media Editors: Scott Sugarman and Kaitlin Coats
Associate Media Editor: Victoria Reuter
Assistant Media Editors: Alex Trivilino and Allison Nicole Smith
Marketing Manager: Ashley Sherwood
Design Director: Rubina Yeh
Art Director: Jillian Burr
Designer: Lisa Buckley
Photo Editor: Ted Szczepanski
Photo Researcher: Lynn Gadson
Permissions Manager: Megan Schindel
Composition/Illustrations: GraphicWorld
Manufacturing: LCS Communications—Kendallville
Permission to use copyrighted material is included in the credits section of this book,
which begins on page C1.
Library of Congress Cataloging-in-Publication Data
Names: Reisberg, Daniel.
Title: Cognition : exploring the science of the mind / Daniel Reisberg, Reed
College.
Description: Seventh Edition. | New York : W. W. Norton & Company, [2018] |
Revised edition of the author’s Cognition, [2016]o | Includes
bibliographical references and index.
Identifiers: LCCN 2018022174 | ISBN 9780393665017 (hardcover)
Subjects: LCSH: Cognitive psychology.
Classification: LCC BF201 .R45 2018 | DDC 153—dc23 LC record available at https://lccn.loc.
gov/2018022174
W. W. Norton & Company, Inc., 500 Fifth Avenue, New York, NY 10110
wwnorton.com
W. W. Norton & Company Ltd., 15 Carlisle Street, London W1D 3BS
1234567890
With love
— always —
for the family that
enriches every
aspect of my life.
Brief Contents
CONTENTS ix
PREFACE xiii
PART 1 THE FOUNDATIONS OF COGNITIVE PSYCHOLOGY
1 The Science of the Mind 2
2 The Neural Basis for Cognition
1
24
PART 2 LEARNING ABOUT THE WORLD AROUND US
61
3 Visual Perception 62
4 Recognizing Objects 106
5 Paying Attention 148
PART 3 MEMORY
193
6 The Acquisition of Memories and the Working-Memory System
7 Interconnections between Acquisition and Retrieval 238
8 Remembering Complex Events 278
PART 4 KNOWLEDGE
323
9 Concepts and Generic Knowledge
10 Language 364
11 Visual Knowledge 410
PART 5 THINKING
194
324
453
12 Judgment and Reasoning 454
13 Problem Solving and Intelligence 498
14 Conscious Thought, Unconscious Thought
Appendix: Research Methods
Glossary G-1
References R-1
Credits C-1
Author Index I-1
Subject Index I-13
546
A-1
vii
Contents
PREFACE
xiii
PART 1THE FOUNDATIONS OF COGNITIVE
PSYCHOLOGY 1
1 The Science of the Mind
2
The Scope of Cognitive Psychology 3 • The Cognitive Revolution 8
• Research in Cognitive Psychology: The Diversity of Methods 17
• Applying Cognitive Psychology 19 • Chapter Review 21
2 The Neural Basis for Cognition
24
Explaining Capgras Syndrome 26 • The Study of the Brain 31
• Sources of Evidence about the Brain 37 • The Cerebral Cortex 44
• Brain Cells 49 • Moving On 55 • Cognitive Psychology and
Education: Food Supplements and Cognition 55 • Chapter Review 58
PART 2
LEARNING ABOUT THE WORLD AROUND US 61
3 Visual Perception
62
The Visual System 64 • Visual Coding 70 • Form Perception 80
• Constancy 87 • The Perception of Depth 92 • Cognitive Psychology
and Education: An “Educated Eye” 99 • Chapter Review 103
4 Recognizing Objects
106
Recognition: Some Early Considerations 110 • Word Recognition 112
• Feature Nets and Word Recognition 116 • Descendants of the
Feature Net 127 • Face Recognition 133 • Top-Down Influences on
Object Recognition 140 • Cognitive Psychology and Education:
Speed-Reading 142 • Chapter Review 145
ix
5 Paying Attention
148
Selective Attention 150 • Selection via Priming 158 • Spatial
Attention 164 • Divided Attention 177 • Practice 183 • Cognitive
Psychology and Education: ADHD 188 • Chapter Review 190
PART 3 MEMORY 193
6The Acquisition of Memories and
the Working-Memory System 194
Acquisition, Storage, and Retrieval 197 • The Route into Memory 198
• A Closer Look at Working Memory 205 • Entering Long-Term
Storage: The Need for Engagement 214 • The Role of Meaning
and Memory Connections 221 • Organizing and Memorizing 224
• The Study of Memory Acquisition 230 • Cognitive Psychology
and Education: How Should I Study? 232 • Chapter Review 235
7Interconnections between Acquisition
and Retrieval 238
Learning as Preparation for Retrieval 241 • Encoding Specificity 244
• The Memory Network 246 • Different Forms of Memory Testing 250
• Implicit Memory 254 • Theoretical Treatments of Implicit Memory 261
• Amnesia 267 • Cognitive Psychology and Education: Familiarity Can
Be Treacherous 273 • Chapter Review 275
8 Remembering Complex Events
278
Memory Errors, Memory Gaps 280 • Memory Errors: A Hypothesis 282
• The Cost of Memory Errors 288 • Avoiding Memory Errors 296
• Forgetting 297 • Memory: An Overall Assessment 302 • Autobiographical
Memory 304 • How General Are the Principles of Memory? 315
• Cognitive Psychology and Education: Remembering for the Long
Term 317 • Chapter Review 320
PART 4 KNOWLEDGE 323
9 Concepts and Generic Knowledge
324
Understanding Concepts 326 • Prototypes and Typicality Effects 329
• Exemplars 334 • The Difficulties with Categorizing via Resemblance
337 • Concepts as Theories 343 • The Knowledge Network 350
• Concepts: Putting the Pieces Together 358 • Cognitive Psychology
and Education: Learning New Concepts 358 • Chapter Review 361
10 Language
364
The Organization of Language 366 • Phonology 368 • Morphemes
and Words 377 • Syntax 378 • Sentence Parsing 382 • Prosody 390
• Pragmatics 391 • The Biological Roots of Language 392 • Language
and Thought 399 • Cognitive Psychology and Education: Writing 404
• Chapter Review 407
x
•
Contents
11 Visual Knowledge
410
Visual Imagery 412 • Chronometric Studies of Imagery 415 • Imagery
and Perception 422 • Visual Imagery and the Brain 424 • Individual
Differences in Imagery 430 • Images Are Not Pictures 435 • Long-Term
Visual Memory 439 • The Diversity of Knowledge 447 • Cognitive
Psychology and Education: Using Imagery 448 • Chapter Review 450
PART 5 THINKING 453
12 Judgment and Reasoning
454
Judgment 456 • Detecting Covariation 463 • Dual-Process Models 466
• Confirmation and Disconfirmation 471 • Logic 476 • Decision
Making 480 • Cognitive Psychology and Education: Making People
Smarter 491 • Chapter Review 494
13 Problem Solving and Intelligence
498
General Problem-Solving Methods 500 • Drawing on Experience 504
• Defining the Problem 509 • Creativity 514 • Intelligence 522
• Intelligence beyond the IQ Test 530 • The Roots of Intelligence 533
• Cognitive Psychology and Education: The Goals of “Education” 539
• Chapter Review 542
14 Conscious Thought, Unconscious Thought
546
The Study of Consciousness 548 • The Cognitive Unconscious 549
• Disruptions of Consciousness 557 • Consciousness and Executive
Control 560 • The Cognitive Neuroscience of Consciousness 566
• The Role of Phenomenal Experience 572 • Consciousness: What Is
Left Unsaid 579 • Cognitive Psychology and Education: Mindfulness
580 • Chapter Review 583
Appendix: Research Methods
A-1
Glossary
G-1
References
R-1
Credits
C-1
Author Index
I-1
Subject Index
I-13
Contents
•
xi
Preface
I
was a college sophomore when I took my first course in cognitive psychology. I was excited about the material then, and, many years later, the
excitement hasn’t faded. Part of the reason lies in the fact that cognitive
psychologists are pursuing fabulous questions, questions that have intrigued
humanity for thousands of years: Why do we think the things we think? Why
do we believe the things we believe? What is “knowledge,” and how secure
(how complete, how accurate) is our knowledge of the world around us?
Other questions asked by cognitive psychologists concern more immediate,
personal, issues: How can I help myself to remember more of the material that
I’m studying in my classes? Is there some better way to solve the problems I
encounter? Why is it that my roommate can study with music on, but I can’t?
And sometimes the questions have important consequences for our social or
political institutions: If an eyewitness reports what he saw at a crime, should we
trust him? If a newspaper raises questions about a candidate’s integrity, how will
voters react?
Of course, we want more than interesting questions—we also want answers
to these questions, and this is another reason I find cognitive psychology so exciting. In the last half-century or so, the field has made extraordinary progress on
many fronts, providing us with a rich understanding of the nature of memory,
the processes of thought, and the content of knowledge. There are many things
still to be discovered—that’s part of the fun. Even so, we already have a lot to say
about all of the questions just posed and many more as well. We can speak to the
specific questions and to the general, to the theoretical issues and to the practical.
Our research has uncovered principles useful for improving the process of education, and we have made discoveries of considerable importance for the criminal
justice system. What I’ve learned as a cognitive psychologist has changed how I
think about my own memory; it’s changed how I make decisions; it’s changed
how I draw conclusions when I’m thinking about events in my life.
On top of all this, I’m also excited about the connections that cognitive
psychology makes possible. In the academic world, intellectual disciplines are
often isolated from one another, sometimes working on closely related problems
xiii
without even realizing it. In the last decades, though, cognitive psychology has
forged rich connections with its neighboring disciplines, and in this book we’ll
touch on topics in philosophy, neuroscience, law and criminal justice, economics, linguistics, politics, computer science, and medicine. These connections bring
obvious benefits, since insights and information can be traded back and forth
between the domains. But these connections also highlight the importance of the
material we’ll be examining, since the connections make it clear that the issues
before us are of interest to a wide range of scholars. This provides a strong signal
that we’re working on questions of considerable power and scope.
I’ve tried in this text to convey all this excitement. I’ve done my best to describe
the questions being asked within my field, the substantial answers we can provide
for these questions, and, finally, some indications of how cognitive psychology is
(and has to be) interwoven with other intellectual endeavors.
I’ve also had other goals in writing this text. In my own teaching, I try to
maintain a balance among many different elements: the nuts and bolts of how
our science proceeds, the data provided by the science, the practical implications
of our research findings, and the theoretical framework that holds all of these
pieces together. I’ve tried to find the same balance in this text. Perhaps most
important, though, I try, both in my teaching and throughout this book, to “tell a
good story,” one that conveys how the various pieces of our field fit together into
a coherent package. Of course, I want the evidence for our claims to be in view,
so that readers can see how our field tests its hypotheses and why our claims must
be taken seriously. But I’ve also put a strong emphasis on the flow of ideas—how
new theories lead to new experiments, and how those experiments can lead to
new theory.
The notion of “telling a good story” also emerges in another way: I’ve always
been impressed by the ways in which the different parts of cognitive psychology
are interlocked. Our claims about attention, for example, have immediate implications for how we can theorize about memory; our theories of object recognition
are linked to our proposals for how knowledge is stored in the mind. Linkages
like these are intellectually satisfying, because they ensure that the pieces of the
puzzle really do fit together. But, in addition, these linkages make the material
within cognitive psychology easier to learn, and easier to remember. Indeed, if I
were to emphasize one crucial fact about memory, it would be that memory is
best when the memorizer perceives the organization and interconnections within
the material being learned. (We’ll discuss this point further in Chapter 6.) With
an eye on this point, I’ve therefore made sure to highlight the interconnections
among various topics, so that readers can appreciate the beauty of our field and
can also be helped in their learning by the orderly nature of our theorizing.
I’ve tried to help readers in other ways, too. First, I’ve tried throughout the
book to make the prose approachable. I want my audience to gain a sophisticated understanding of the material in this text, but I don’t want readers to
struggle with the ideas.
Second, I’ve taken various steps that I hope will foster an “alliance” with
readers. My strategy here grows out of the fact that, like most teachers, I value
the questions I receive from students and the discussions I have with them. In
xiv •
Preface
the classroom, this allows a two-way flow of information that unmistakably
improves the educational process. Of course, a two-way flow isn’t possible in
a textbook, but I’ve offered what I hope is a good approximation: Often, the
questions I hear from students, and the discussions I have with them, focus
on the relevance of the course material to students’ own lives, or relevance to
the world outside of academics. I’ve tried to capture that dynamic, and to present
my answers to these student questions, in the essay at the end of each chapter
(I’ll say more about these essays in a moment). These essays appear under the
banner Cognitive Psychology and Education, and—as the label suggests—
the essays will help readers understand how the materials covered in that chapter
matter for (and might change!) the readers’ own learning. In addition, I’ve written a separate series of essays (available online), titled Cognitive Psychology and
the Law, to explore how each chapter’s materials matter in another arena—the
enormously important domain of the justice system. I hope that both types of
essays—Education and Law—help readers see that all of this material is indeed
relevant to their lives, and perhaps as exciting for them as it is for me.
Have I met all of these goals? You, the readers, will need to be the judges of
this. I would love to hear from you about what I’ve done well in the book and
what I could have done better; what I’ve covered (but should have omitted) and
what I’ve left out. I’ll do my best to respond to every comment. You can reach me
via email (reisberg@reed.edu); I’ve been delighted to get comments from readers
about previous editions, and I hope for more emails with this edition.
An Outline of the Seventh Edition
The book’s 14 chapters are designed to cover the major topics within cognitive psychology. The chapters in Part 1 lay the foundation. Chapter 1 provides the conceptual and historical background for the subsequent chapters.
In addition, this chapter seeks to convey the extraordinary scope of the field
and why, therefore, research on cognition is so important. The chapter also
highlights the relationship between theory and evidence in cognitive psychology, and it discusses the logic on which this field is built.
Chapter 2 then offers a brief introduction to the study of the brain. Most of
cognitive psychology is concerned with the functions that our brains make possible, and not the brain itself. Nonetheless, our understanding of cognition has
certainly been enhanced by the study of the brain, and throughout this book
we’ll use biological evidence as one means of evaluating our theories. Chapter 2
is designed to make this evidence fully accessible to the reader—by providing
a quick survey of the research tools used in studying the brain, an overview of
the brain’s anatomy, and also an example of how we can use brain evidence as a
source of insight into cognitive phenomena.
Part 2 of the book considers the broad issue of how we gain information from
the world. Chapter 3 covers visual perception. At the outset, this chapter links to
the previous (neuroscience) chapter with descriptions of the eyeball and the basic
mechanisms of early visual processing. In this context, the chapter introduces
the crucial concept of parallel processing and the prospect of mutual influence
Preface
•
xv
among separate neural mechanisms. From this base, the chapter builds toward a
discussion of the perceiver’s activity in shaping and organizing the visual world,
and explores this point by discussing the rich topics of perceptual constancy and
perceptual illusions.
Chapter 4 discusses how we recognize the objects that surround us. This
seems a straightforward matter—what could be easier than recognizing a telephone, or a coffee cup, or the letter Q? As we’ll see, however, recognition is
surprisingly complex, and discussion of this complexity allows me to amplify key
themes introduced in earlier chapters: how active people are in organizing and
interpreting the information they receive from the world; the degree to which
people supplement the information by relying on prior experience; and the ways
in which this knowledge can be built into a network.
Chapter 5 then considers what it means to “pay attention.” The first half of
the chapter is concerned largely with selective attention—cases in which you seek
to focus on a target while ignoring distractors. The second half of the chapter is
concerned with divided attention (“multi-tasking”)—that is, cases in which you
seek to focus on more than one target, or more than one task, at the same time.
Here, too, we’ll see that seemingly simple processes turn out to be more complicated than one might suppose.
Part 3 turns to the broad topic of memory. Chapters 6, 7, and 8 start with a
discussion of how information is “entered’’ into long-term storage, but then turn
to the complex interdependence between how information is first learned and
how that same information is subsequently retrieved. A recurrent theme in this
section is that learning that’s effective for one sort of task, one sort of use, may
be quite ineffective for other uses. This theme is examined in several contexts,
and leads to a discussion of research on unconscious memories—so-called memory without awareness. These chapters also offer a broad assessment of human
memory: How accurate are our memories? How complete? How long-lasting?
These issues are pursued both with regard to theoretical treatments of memory
and also with regard to the practical consequences of memory research, including
the application of this research to the assessment, in the courtroom, of eyewitness
testimony.
The book’s Part 4 is about knowledge. Earlier chapters show over and over
that humans are, in many ways, guided in their thinking and experiences by what
they already know—that is, the broad pattern of knowledge they bring to each
new experience. This invites the questions posed by Chapters 9, 10, and 11: What
is knowledge? How is it represented in the mind? Chapter 9 tackles the question
of how “concepts,” the building blocks of our knowledge, are represented in the
mind. Chapters 10 and 11 focus on two special types of knowledge. Chapter 10
examines our knowledge about language; Chapter 11 considers visual knowledge and examines what is known about mental imagery.
The chapters in Part 5 are concerned with the topic of thinking. Chapter 12
examines how each of us draws conclusions from evidence—including cases
in which we are trying to be careful and deliberate in our judgments, and also
cases of informal judgments of the sort we often make in our everyday lives. The
chapter then turns to the question of how we reason from our beliefs—how we
xvi •
Preface
check on whether our beliefs are correct, and how we draw conclusions, based on
things we already believe. The chapter also considers the practical issue of how
errors in thinking can be diminished through education.
Chapter 13 is also about thinking, but with a different perspective: This
chapter considers some of the ways people differ from one another in their ability to solve problems, in their creativity, and in their intelligence. The chapter
also addresses the often heated, often misunderstood debate about how different
groups—especially American Whites and African Americans—might (or might
not) differ in their intellectual capacities.
The final chapter in the book does double service. First, it pulls together
many of the strands of contemporary research relevant to the topic of
consciousness—what consciousness is, and what consciousness is for. In addition, most readers will reach this chapter at the end of a full semester’s work,
a point at which they are well served by a review of the topics already covered
and ill served by the introduction of much new material. Therefore, this chapter
draws many of its themes and evidence from previous chapters, and in that
fashion it serves as a review of points that appear earlier in the book. Chapter
14 also highlights the fact that we’re using these materials to approach some of
the greatest questions ever asked about the mind, and, in that way, this chapter
should help to convey some of the power of the material we’ve been discussing
throughout the book.
New in the Seventh Edition
What’s new in this edition? Every chapter contains new material, in most
cases because readers specifically requested the new content! Chapter 1, for
example, now includes discussion of how the field of cognitive psychology
emerged in the 1950s and 1960s. Chapter 4 includes coverage of recent work
on how people differ from one another in their level of face-recognition skill.
Chapter 5 discusses what it is that people pay attention to, with a quick
summary of research on how men and women differ in what they focus on,
and how different cultures seem to differ in what they focus on. Chapter 8
discusses a somewhat controversial and certainly dramatic study showing
that college students can be led to a false memory of a time they committed a
felony (an armed assault) while in high school; this chapter also now includes
coverage of the social nature of remembering. Chapter 10 now discusses the
topics of prosody and pragmatics. Chapter 12 discusses the important difference between “opt-in” and “opt-out” procedures for social policy, and
Chapter 14 now includes discussion of (both the myths and the reality of)
subliminal perception.
In this edition, I’ve also added three entirely new features. First, my students
are always curious to learn how cognitive psychology research can be applied to
issues and concerns that arise in everyday life. I’ve therefore added a Cognition
Outside the Lab essay to every chapter. For example, in Chapter 4, in discussing how word recognition proceeds, I’ve tackled the question of how the choice
of font can influence readers (sometimes in good ways and sometimes not). In
Preface
•
xvii
Chapter 7, I’ve written about cryptoplagiarism, a pattern in which you can steal
another person’s ideas without realizing it!
Second, I have always believed that, as someone teaching cognitive psychology, I need to respect the practical lessons of my field. As one example, research
suggests that students’ understanding and memory are improved if they pause
and reflect on materials they’ve just heard in a lecture or just read in a book.
“What did I just hear? What were the main points? Which bits were new, and
which bits had I thought about before?” Guided by that research, I’ve added Test
Yourself questions throughout the book. These questions are then echoed at the
end of the chapter, with the aim of encouraging readers to do another round of
reflection. All these questions are designed to be easy and straightforward—but
should, our research tells us, be genuinely helpful for readers.
Third, the topics covered in this book have many implications, and I hope
readers will find it both fun and useful to think about some of these implications.
On this basis, every chapter also ends with a couple of Think About It questions,
inviting readers to extend the chapter’s lessons into new territory. For example,
at the end of Chapter 3, I invite readers to think about how research on attention
might help us understand what happens in the focused exercise of meditation
(including Buddhist meditation). The question at the end of Chapter 7 invites
readers to think through how we might explain the eerie sensation of déjà vu.
A question at the end of Chapter 8 explores how your memory is worse than a
video recorder, and also how it’s better than a video recorder.
Other Special Features
In addition, I have (of course) held on to features that were newly added in
the previous edition—including an art program that showcases the many
points of contact between cognitive psychology and cognitive neuroscience,
and the “What if . . .” section that launches each chapter. The “What if . . .”
material serves several aims. First, the mental capacities described in each
chapter (the ability to recognize objects, the ability to pay attention, and so
on) are crucial for our day-to-day functioning, and to help readers understand this point, most of the “What if . . .” sections explore what someone’s
life is like if they lack the relevant capacity. Second, the “What if . . .” sections
are rooted in concrete, human stories; they talk about specific individuals
who lack these capacities. I hope these stories will be inviting and thoughtprovoking for readers, motivating them to engage the material in a richer
way. And, third, most of the “What if . . .” sections involve people who have
lost the relevant capacity through some sort of brain damage. These sections
therefore provide another avenue through which to highlight the linkage
between cognitive psychology and cognitive neuroscience.
This edition also includes explicit coverage of Research Methods. As in the
previous edition, this material is covered in an appendix, so that it’s easily accessible to all readers, but set to the side to accommodate readers (or instructors)
who prefer to focus on the book’s substantive content. The appendix is divided
into separate essays for each chapter, so that the appendix can be used on a
xviii •
Preface
chapter-by-chapter basis. This organization will help readers see, for each
chapter, how the research described in the chapter unfolds, and it will simultaneously provide a context for each methods essay so that readers can see why the
methods are so important.
The appendix is surely no substitute for a research methods course, but nonetheless it’s sequenced in a manner that builds toward a broad understanding of
how the scientific method plays out in our field. An early essay, for example,
works through the question of what a “testable hypothesis” is, and why this is
so important; another essay works through the power of random assignment;
another discusses how we deal with confounds. In all cases, my hope is that the
appendix will guide readers toward a sophisticated understanding of why our
research is as it is, and why, therefore, our research is so persuasive.
I have already mentioned the end-of-chapter essays on Cognitive Psychology
and Education, which show students how cognitive psychology is connected to
their own learning. Readers often seek “take-home messages” from the material
that will, in a direct way, benefit them. We are, after all, talking about memory, and
students obviously are engaged in an endeavor of putting lots of new information—
information they’re learning in their courses—into their memories. We’re talking
about attention, and students often struggle with the chore of keeping themselves
“on task” and “on target.” In light of these points, the end-of-chapter essays build
a bridge between the content in the chapter and the concerns that fill students’
lives. This will, I hope, make the material more useful for students, and also make
it clear just how important an enterprise cognitive psychology is.
There are also essays in the ebook on Cognitive Psychology and the Law.
Here, I’ve drawn on my own experience in working with law enforcement and
the criminal justice system. In this work, I’m called on to help juries understand
how an eyewitness might be certain in his recollection, but mistaken. I also work
with police officers to help them elicit as much information from a witness as
possible, without leading the witness in any way. Based on this experience, the
online essays discuss how the material in each chapter might be useful for the
legal system. These essays will, I hope, be immediately interesting for readers, and
will also make it obvious why it’s crucial that the science be done carefully and
well—so that we bring only high-quality information into the legal system.
I’ve also included Demonstrations in the ebook to accompany the book’s
description of key concepts in the field. Many of these demos are miniature versions of experimental procedures, allowing students to see for themselves what
these experiments involve, and also allowing them to see just how powerful
many of our effects are. Readers who want to run the demos for themselves as
they read along certainly can; instructors who want to run the demos within
their classrooms (as I sometimes do) are certainly encouraged to do so. Instructors who want to use the demos in discussion sections, aside from the main
course, can do that as well. In truth, I suspect that some demos will work better
in one of these venues, and that other demos will work better in others; but,
in all cases, I hope the Demonstrations help bring the material to life—putting
students directly in contact with both our experimental methods and our experimental results.
Preface
•
xix
As in previous editions, this version of Cognition also comes with various
supplementary materials, some aimed at students, and some aimed at instructors.
For Students
ZAPS Cognition Labs. Every copy of the text comes packaged with free
access to ZAPS Cognition Labs, an updated revision of Norton’s popular
online psychology labs. Crafted specifically to support cognitive psychology
courses, this version helps students learn about core psychological phenomena. Each lab (one or two per chapter) begins with a brief introduction that
relates the topic to students’ lives. Students then engage in a hands-on experience that, for most labs, produces data based on their individual responses.
The theories behind the concepts are then explained alongside the data the
student has generated. Also, an assessment component lets students confirm
that they understand the concepts central to each lab. Finally, this edition
of Cognition is accompanied by five new ZAPS labs: Encoding Specificity,
Mental Rotation 3D, Memory Span, Operation Span, and Selective Attention.
Ebook. Every print copy of the text comes packaged with free access to the
ebook. The ebook can also be purchased separately at a fraction of the price of
the printed version. The ebook has several advantages over the print text. First,
the ebook includes Demonstrations—quick, pen-and-paper mini experiments—
designed to show students what key experiments involve and how powerful
many of the effects are. Second, the ebook includes the Cognitive Psychology and
the Law essays, described above. In addition, the ebook can be viewed on any
device—laptop, tablet, phone, public computer—and will stay synced between
devices. The ebook is therefore a perfect solution for students who want to learn
in more convenient settings—and pay less for doing so.
For Instructors
All instructor resources for this edition of Cognition can be accessed via the
“Instructor Resources” tile at the following URL: https://digital.wwnorton
.com/cognition7.
Interactive Instructor’s Guide (IIG). This online repository of teaching assets
offers material for every chapter that both veteran and novice instructors of the
course will find helpful. Searchable by chapter or asset class, the IIG provides
multiple resources for teaching: links to online video clips (selected and annotated by the author), teaching suggestions, and other class activities and exercises.
It also includes all of the Education, Law, and Research Methods essays described
above, as well as discussion questions to support the Education and Law essays.
The demonstrations from the ebook can also be found here. This repository of
lecture and teaching materials functions both as a course prep tool and as a
means of tracking the latest ideas in teaching the cognitive psychology course.
I’m especially excited about the online video clips. Students love videos and
probably spend more time than they should surfing the Internet (and YouTube
in particular) for fun clips. As it turns out, though, YouTube contains far more
xx
•
Preface
than cute-kittens movies; it also contains intriguing, powerful material directly
relevant to the topics in this text. The IIG therefore provides a listing of carefully selected online videos to accompany each of the chapters. (A dozen of these
videos are newly added for the seventh edition!) The annotated list describes each
clip, and gives information about timing, in ways that should make these videos
easy to use in the classroom. I use them in my own teaching, and my students
love them. But let me also make a request: I’m sure there are other videos available that I haven’t seen yet. I’ll therefore be grateful to any readers who help me
broaden this set, so that we can make this resource even better.
Test Bank. The test bank features over 900 questions, including multiplechoice and short-answer questions for each chapter. I have personally vetted each
question, and all questions have been updated according to Norton’s assessment
guidelines to make it easy for instructors to construct quizzes and exams that are
meaningful and diagnostic. All questions are classified according to learning objective, text section, difficulty, and question type. This Norton test bank is available
with ExamView Test Generator software, allowing instructors to create, administer, and manage assessments. The intuitive test-making wizard makes it easy to
create customized exams. Other features include the ability to create paper exams
with algorithmically generated variables and to export files directly to your LMS.
Lecture PowerPoints. These text-focused PowerPoints follow the chapter
outlines, include figures from the text, and feature instructor-only notes.
Art Slides. All the figures, photos, and tables from the text are offered as
JPEGs, both separately and embedded in a PowerPoint for each chapter. All text
art is enhanced for optimal viewing when projected in large classrooms.
Coursepack (Blackboard, Canvas, Angel, Moodle, and other LMS systems).
Available at no cost to professors or students, Norton coursepacks for online,
hybrid, or lecture courses are available in a variety of formats. With a simple
download from the instructor’s website, an adopter can bring high-quality Norton
digital media into a new or existing online course (no extra student passwords
required), and it’s theirs to keep. Instructors can edit assignments at the question level and set up custom grading policies to assess student understanding. In
addition to the instructor resources listed above, the coursepack includes additional chapter quizzes, flashcards, chapter outlines, chapter summaries, all of the
Education, Law, and Research Methods essays described above, and additional
questions on the essays.
Acknowledgments
Finally, let me turn to the happiest of chores—thanking all of those who have
contributed to this book. I begin with those who helped with the previous editions: Bob Crowder (Yale University) and Bob Logie (University of Aberdeen)
both read the entire text of the first edition, and the book was unmistakably improved by their insights. Other colleagues read, and helped me enormously with, specific chapters: Enriqueta Canseco-Gonzalez (Reed College),
Rich Carlson (Pennsylvania State University), Henry Gleitman (University
of Pennsylvania), Lila Gleitman (University of Pennsylvania), Peter Graf
Preface
•
xxi
(University of British Columbia), John Henderson (Michigan State University),
Jim Hoffman (University of Delaware), Frank Keil (Cornell University), Mike
McCloskey (Johns Hopkins University), Hal Pashler (UCSD), Steve Pinker
(MIT), and Paul Rozin (University of Pennsylvania).
The second edition was markedly strengthened by the input and commentary
provided by Martin Conway (University of Bristol), Kathleen Eberhard (Notre
Dame University), Howard Egeth (Johns Hopkins University), Bill Gehring
(University of Michigan), Steve Palmer (University of California, Berkeley), Henry
Roediger (Washington University), and Eldar Shafir (Princeton University).
In the third edition, I was again fortunate to have the advice, criticism, and
insights provided by a number of colleagues who, together, made the book
better than it otherwise could have been, and I’d like to thank Rich Carlson (Penn
State), Richard Catrambone (Georgia Tech), Randall Engle (Georgia Tech), Bill
Gehring and Ellen Hamilton (University of Michigan), Nancy Kim (Rochester
Institute of Technology), Steve Luck (University of Iowa), Michael Miller
(University of California, Santa Barbara), Evan Palmer, Melinda Kunar, and
Jeremy Wolfe (Harvard University), Chris Shunn (University of Pittsburgh), and
Daniel Simons (University of Illinois).
A number of colleagues also provided their insights and counsel for the fourth
edition. I’m therefore delighted to thank Ed Awh (University of Oregon), Glen
Bodner (University of Calgary), William Gehring (University of Michigan),
Katherine Gibbs (University of California, Davis), Eliot Hazeltine (University of
Iowa), William Hockley (Wilfrid Laurier University), James Hoffman (University
of Delaware), Helene Intraub (University of Delaware), Vikram Jaswal (University
of Virginia), Karsten Loepelmann (University of Alberta), Penny Pexman
(University of Calgary), and Christy Porter (College of William and Mary).
Then, even more people to thank for their help with the fifth edition: Karin M.
Butler (University of New Mexico), Mark A. Casteel (Penn State University, York),
Alan Castel (University of California, Los Angeles), Robert Crutcher (University
of Dayton), Kara D. Federmeier (University of Illinois, Urbana-Champaign),
Jonathan Flombaum (Johns Hopkins University), Katherine Gibbs (University of
California, Davis), Arturo E. Hernandez (University of Houston), James Hoeffner
(University of Michigan), Timothy Jay (Massachusetts College of Liberal Arts),
Timothy Justus (Pitzer College), Janet Nicol (University of Arizona), Robyn T.
Oliver (Roosevelt University), Raymond Phinney (Wheaton College, and his
comments were especially thoughtful!), Brad Postle (University of Wisconsin,
Madison), Erik D. Reichle (University of Pittsburgh), Eric Ruthruff (University of
New Mexico), Dave Sobel (Brown University), Martin van den Berg (California
State University, Chico), and Daniel R. VanHorn (North Central College).
For the sixth edition: Michael Dodd (University of Nebraska, Lincoln),
James Enns (University of British Columbia), E. Christina Ford (Penn State
University), Danielle Gagne (Alfred University), Marc Howard (Boston University),
B. Brian Kuhlman (Boise State University), Guy Lacroix (Carleton University), Ken
Manktelow (University of Wolverhampton), Aidan Moran (University College
Dublin, Ireland), Joshua New (Barnard College), Janet Nicol (University of
xxii •
Preface
Arizona), Mohammed K. Shakeel (Kent State University), David Somers (Boston
University), and Stefan Van der Stigchel (Utrecht University).
And now, happily, for the current edition: Alan Castel (UCLA), Jim Hoelzle
(Marquette University), Nate Kornell (Williams College), Catherine Middlebrooks (UCLA), Stefanie Sharman (Deakin University), Erin Sparck (UCLA), and
Cara Laney Thede (The College of Idaho).
I also want to thank the people at Norton. I’ve had a succession of terrific
editors, and I’m grateful to Ken Barton, Sheri Snavely, Aaron Javsicas, and Jon
Durbin for their support and fabulous guidance over the years. There’s no question that the book is stronger, clearer, and better because of their input and advice.
I also want to thank David Bradley for doing a fabulous job of keeping this
project on track, and also Eve Sanoussi and Katie Pak for their extraordinary
work in helping me bring out a book of the highest quality. (I should also thank
them for putting up with my gruff impatience when things don’t work quite
as we all hoped.) I’m also delighted with the design that Rubina Yeh, Jillian
Burr, and Lisa Buckley created; Ted Szczepanski has been great in dealing with
my sometimes-zany requests for photographs. Thanks in advance to Ashley
Sherwood and the Norton sales team; I am, of course, deeply grateful for all you
do. Thanks also to the crew that has produced the ZAPS, the Internet presence,
and the various supplements for this book.
And I once again get to celebrate the pleasure of working with Alice Vigliani
as manuscript editor. Alice is, of course, a terrific editor, but, in addition, she’s efficient and fun to work with. As I’ve said before: If she tends her peonies with the
same skill that she devotes to her editing, her garden must be marvelous indeed.
Finally, it brings me joy to reiterate with love the words I said in the previous
edition: In countless ways, Friderike makes all of this possible and worthwhile.
She forgives me the endless hours at the computer, tolerates the tension when
I’m feeling overwhelmed by deadlines, and is always ready to read my pages and
offer thoughtful, careful, instructive insights. My gratitude to, and love for, her
are boundless.
Daniel Reisberg
Portland, Oregon
Preface
•
xxiii
The Foundations
of Cognitive
Psychology
part
1
W
hat is cognitive psychology? In Chapter 1, we’ll define this discipline and offer a sketch of what this field can teach us — through
its theories and its practical applications. We’ll also provide
a brief history to explain why cognitive psychology has taken the form that
it has.
Chapter 2 has a different focus. At many points in this book, we’ll draw
insights from the field of cognitive neuroscience — the effort toward understanding our mental functioning through close study of the brain and nervous
system. To make sure this biological evidence is useful, though, we need to
provide some background, and that’s the main purpose of Chapter 2. There,
we’ll offer a rough mapping of what’s where in the brain, and we’ll describe
the functioning of many of the brain’s parts. We’ll also discuss what it means
to describe the functioning of this or that brain region, because, as we will see,
each of the brain’s parts is highly specialized in what it does. As a result, mental
achievements such as reading, remembering, or deciding depend on the coordinated functioning of many different brain regions, with each contributing its
own small bit to the overall achievement.
1
1
chapter
The Science of
the Mind
Almost everything you do, and everything you feel or say, depends on
your cognition — what you know, what you remember, and what you think.
As a result, the book you’re now reading — a textbook on cognition —
describes the foundation for virtually every aspect of who you are.
As illustrations of this theme, in a few pages we’ll consider the way
in which your ability to cope with grief depends on how your memory
functions. We’ll also discuss the role that memory plays in shaping your
self-image — and, therefore, your self-esteem. As another example, we’ll
discuss a case in which your understanding of a simple story depends on
the background knowledge that you supply. Examples like these make it
clear that cognition matters in an extraordinary range of circumstances,
and it’s on this basis that our focus in this book is on the intellectual
foundation of almost every aspect of human experience.
The Scope of Cognitive Psychology
When the field of cognitive psychology was first launched, it was broadly
focused on the scientific study of knowledge, and this focus led immediately
to a series of questions: How is knowledge acquired? How is knowledge
retained so that it’s available when needed? How is knowledge used — whether
as a basis for making decisions or as a means of solving problems?
These are great questions, and it’s easy to see that answering them might
be quite useful. For example, imagine that you’re studying for next Wednesday’s exam, but for some reason the material just won’t “stick” in your
memory. You find yourself wishing, therefore, for a better strategy to use in
studying and memorizing. What would that strategy be? Is it possible to have
a “better memory”?
As a different case, let’s say that while you’re studying, your friend is moving around in the room, and you find this quite distracting. Why can’t you
just shut out your friend’s motion? Why don’t you have better control over
your attention and your ability to concentrate?
Here’s one more example: You’re looking at your favorite Internet news
site, and you’re horrified to learn how many people have decided to vote for
candidate X. How do people decide whom to vote for? For that matter, how
do people decide what college to attend, or which car to buy, or even what
to have for dinner? And how can we help people make better decisions — so
that, for example, they choose healthier foods, or vote for the candidate who
(in your view) is preferable?
3
preview of chapter themes
•
•
he chapter begins with a description of the scope of
T
cognitive psychology. The domain of this field includes
activities that are obviously “intellectual” (such as remembering, paying attention, or making judgments) but also
a much broader range of activities that depend on these
intellectual achievements.
hat form should a “science of the mind” take? We disW
cuss the difficulties in trying to study the mind by means
of direct observation. But we also explore why we must
study the mental world if we’re to understand behavior;
the reason is that our behavior depends in crucial ways on
how we perceive and understand the world around us.
•
ombining these themes, we come to the view that we
C
must study the mental world indirectly. But as we will see,
the method for doing this is the method used by most
sciences.
Before we’re through, we’ll consider evidence pertinent to all of these
questions. Let’s note, though, that in these examples, things aren’t going
as you might have wished: You remember less than you want to; you can’t
ignore a distraction; the voters make a choice you don’t like. What about
the other side of the picture? What about the remarkable intellectual feats
that humans achieve — brilliant deductions or creative solutions to complex
problems? In this text, we’ll also discuss these cases and explore how people
manage to accomplish the great things they do.
CELEBRATING HUMAN ACHIEVEMENTS
Many of the text’s examples involve failures or limitations in our cognition. But we
also need to explain the incredible intellectual achievements of our species — the
complex problems we’ve solved and the extraordinary devices we’ve invented.
4
•
C H A P T E R O N E The Science of the Mind
The Broad Role for Memory
The questions we’ve mentioned so far might make it sound like cognitive
psychology is concerned just with your functioning as an intellectual —
your ability to remember, or to pay attention, or to think through options
when making a choice. As we’ve said, though, the relevance of cognitive
psychology is much broader — thanks to the fact that a huge range of your
actions, thoughts, and feelings depend on your cognition. As one way to
convey this point, let’s ask: When we investigate how memory functions,
what’s at stake? Or, to turn this around, what aspects of your life depend
on memory?
You obviously rely on memory when you’re taking an exam — memory
for what you learned during the term. Likewise, you rely on memory when
you’re at the supermarket and trying to remember the cheesecake recipe so
that you can buy the right ingredients. You also rely on memory when you’re
reminiscing about childhood. But what else draws on memory?
Consider this simple story (adapted from Charniak, 1972):
Betsy wanted to bring Jacob a present. She shook her piggy bank. It
made no sound. She went to look for her mother.
This four-sentence tale is easy to understand, but only because you provided
important bits of background. For example, you weren’t at all puzzled about
why Betsy was interested in her piggy bank; you weren’t puzzled, specifically,
about why the story’s first sentence led naturally to the second. This is
because you already knew (a) that the things one gives as presents are often
things bought for the occasion (rather than things already owned), (b) that
buying things requires money, and (c) that money is sometimes stored in
piggy banks. Without these facts, you would have wondered why a desire to
give a gift would lead someone to her piggy bank. (Surely you didn’t think
Betsy intended to give the piggy bank itself as the present!) Likewise, you
immediately understood why Betsy shook her piggy bank. You didn’t suppose
that she was shaking it in frustration or trying to find out if it would make a
good percussion instrument. Instead, you understood that she was trying to
determine its contents. But you knew this fact only because you already knew
(d) that Betsy was a child (because few adults keep their money in piggy
banks), (e) that children don’t keep track of how much money is in their
banks, and (f) that piggy banks are made out of opaque material (and so
a child can’t simply look into the bank to see what’s inside). Without these
facts, Betsy’s shaking of the bank would make no sense. Similarly, you understood what it meant that the bank made no sound. That’s because you know
(g) that it’s usually coins (not bills) that are kept in piggy banks, and (h) that
coins make noise when they’re shaken. If you didn’t know these facts, you
might have interpreted the bank’s silence, when it was shaken, as good news,
indicating perhaps that the bank was jammed full of $20 bills — an inference
that would have led you to a very different expectation for how the story
would unfold from there.
A SIMPLE STORY
What is involved in your
understanding of this simple
story? Betsy wanted to bring
Jacob a present. She shook
her piggy bank. It made no
sound. She went to look for
her mother.
The Scope of Cognitive Psychology
•
5
TRYING TO FOCUS
Often, you want to focus on
just one thing, and you want
to shut out the other sights
and sounds that are making
it hard for you to concentrate. What steps should you
take to promote this focus
and to avoid distraction?
6 •
Of course, there’s nothing special about the “Betsy and Jacob” story,
and we’d uncover a similar reliance on background knowledge if we
explored how you understand some other narrative, or follow a conversation, or comprehend a TV show. Our suggestion, in other words, is that
many (perhaps all) of your encounters with the world depend on your
supplementing your experience with knowledge that you bring to the situation.
And perhaps this has to be true. After all, if you didn’t supply the relevant
bits of background, then anyone telling the “Betsy and Jacob” story would
need to spell out all the connections and all the assumptions. That is, the
story would have to include all the facts that, with memory, are supplied
by you. As a result, the story would have to be much longer, and the telling of it much slower. The same would be true for every story you hear,
every conversation you participate in. Memory is thus crucial for each of
these activities.
Amnesia and Memory Loss
Here is a different sort of example: In Chapter 7, we will consider cases of
clinical amnesia — cases in which someone, because of brain damage, has lost
the ability to remember certain materials. These cases are fascinating at many
levels and provide key insights into what memory is for. Without memory,
what is disrupted?
H.M. was in his mid-20s when he had brain surgery intended to control
his severe epilepsy. The surgery was, in a narrow sense, a success, and H.M.’s
epilepsy was brought under control. But this gain came at an enormous cost,
because H.M. essentially lost the ability to form new memories. He survived
for more than 50 years after the operation, and for all those years he had
little trouble remembering events prior to the surgery. But H.M. seemed completely unable to recall any event that occurred after his operation. If asked
who the president is, or about recent events, he reported facts and events that
were current at the time of the surgery. If asked questions about last week, or
even an hour ago, he recalled nothing.
This memory loss had massive consequences for H.M.’s life, and some
of the consequences are surprising. For example, he had an uncle he was
very fond of, and he occasionally asked his hospital visitors how his uncle
was doing. Unfortunately, the uncle died sometime after H.M.’s surgery, and
H.M. was told this sad news. The information came as a horrible shock, but
because of his amnesia, H.M. soon forgot about it.
Sometime later, because he’d forgotten about his uncle’s death, H.M. again
asked how his uncle was doing and was again told of the death. But with no
memory of having heard this news before, he was once more hearing it “for
the first time,” with the shock and grief every bit as strong as it was initially.
Indeed, each time he heard this news, he was hearing it “for the first time.”
With no memory, he had no opportunity to live with the news, to adjust to it.
As a result, his grief could not subside. Without memory, H.M. had no way
to come to terms with his uncle’s death.
C H A P T E R O N E The Science of the Mind
H.M.’S BRAIN
When H.M. died in 2008, the world learned his full name — Henry Molaison. Throughout his life, H.M. had cooperated with researchers in many studies of his memory
loss. Even after his death, H.M. is contributing to science: His brain (shown here) was
frozen and has now been sliced into sections for detailed anatomical study. Unfortunately, though, there has been debate over who “owns” H.M.’s brain and how we
might interpret some observations about his brain (see, for example, Dittrich, 2016).
A different glimpse of memory function comes from some of H.M.’s
comments about what it felt like to be in his situation. Let’s start here with
the notion that for those of us without amnesia, numerous memories support our conception of who we are: We know whether we deserve praise for
our good deeds or blame for our transgressions because we remember those
good deeds and transgressions. We know whether we’ve kept our promises
or achieved our goals because, again, we have the relevant memories. None
of this is true for people who suffer from amnesia, and H.M. sometimes commented that in important ways, he didn’t know who he was. He didn’t know
if he should be proud of his accomplishments or ashamed of his crimes; he
didn’t know if he’d been clever or stupid, honorable or dishonest, industrious or lazy. In a sense, then, without a memory, there is no self. (For broader
discussion, see Conway & Pleydell-Pearce, 2000; Hilts, 1995.)
What, then, is the scope of cognitive psychology? As we mentioned earlier,
this field is sometimes defined as the scientific study of the acquisition, retention, and use of knowledge. We’ve now seen, though, that “knowledge” (and
hence the study of how we gain and use knowledge) is relevant to a huge
range of concerns. Our self-concept, it seems, depends on our knowledge
(and, in particular, on our memory for various episodes in our past). Our
The Scope of Cognitive Psychology
•
7
TEST YOURSELF
1.Why is memory crucial
for behaviors and
mental operations that
don’t in any direct or
explicit way ask you
“to remember”?
2. What aspects of H.M.’s
life were disrupted as
a result of his amnesia?
emotional adjustments to the world rely on our memories. Even our ability to understand a simple story — or, presumably, our ability to understand
any experience — depends on our supplementing that experience with some
knowledge.
The suggestion, then, is that cognitive psychology can help us understand
capacities relevant to virtually every moment of our lives. Activities that don’t
appear to be intellectual would collapse without the support of our cognitive functioning. The same is true whether we’re considering our physical
movements through the world, our social lives, our emotions, or any other
domain. This is the scope of cognitive psychology and, in a real sense, the
scope of this book.
The Cognitive Revolution
The enterprise that we now call “cognitive psychology” is a bit more than
50 years old, and the emergence of this field was in some ways dramatic.
Indeed, the science of psychology went through a succession of changes in
the 1950s and 1960s that are often referred to as psychology’s “cognitive revo­
lution.” This “revolution” involved a new style of research, aimed initially
at questions we’ve already met: questions about memory, decision making,
and so on. But this new type of research, and its new approach to theorizing,
soon influenced other domains, with the result that the cognitive revolution
dramatically changed the intellectual map of our field.
The cognitive revolution centered on two key ideas. One idea is that
the science of psychology cannot study the mental world directly. A second
idea is that the science of psychology must study the mental world if we’re
going to understand behavior. As a path toward understanding these ideas,
let’s look at two earlier traditions in psychology that offered a rather different perspective. Let’s emphasize, though, that our purpose here is not
to describe the full history of modern cognitive psychology. That history
is rich and interesting, but our goal is a narrow one — to explain why the
cognitive revolution’s themes were as they were. (For readers interested in
the history, see Bartlett, 1932; Benjamin, 2008; Broadbent, 1958; Malone,
2009; Mandler, 2011.)
The Limits of Introspection
In the late 19th century, Wilhelm Wundt (1832–1920) and his student
Edward Bradford Titchener (1867–1927) launched a new research enterprise,
and according to many scholars it was their work that eventually led to the
modern field of experimental psychology. In Wundt’s and Titchener’s view,
psychology needed to focus largely on the study of conscious mental events —
feelings, thoughts, perceptions, and recollections. But how should these events
be studied? These early researchers started with the fact that there is no way
for you to experience my thoughts, or I yours. The only person who can experience or observe your thoughts is you. Wundt, Titchener, and their colleagues
8 •
C H A P T E R O N E The Science of the Mind
WILHELM WUNDT
Wilhelm Wundt (1832–1920)
is shown here sitting and
surrounded by his colleagues
and students. Wundt is often
regarded as the “father of
experimental psychology.”
concluded, therefore, that the only way to study thoughts is through introspection, or “looking within,” to observe and record the content of our own
mental lives and the sequence of our own experiences.
Wundt and Titchener insisted, though, that this introspection could not
be casual. Instead, introspectors had to be meticulously trained: They were
given a vocabulary to describe what they observed; they were taught to be as
careful and as complete as possible; and above all, they were trained simply
to report on their experiences, with a minimum of interpretation.
This style of research was enormously influential for several years, but
psychologists gradually became disenchanted with it, and it’s easy to see
why. As one concern, these investigators soon had to acknowledge that
some thoughts are unconscious, which meant that introspection was limi­ted
as a research tool. After all, by its very nature introspection is the study
of conscious experiences, so of course it can tell us nothing about unconscious events.
Indeed, we now know that unconscious thought plays a huge part in
our mental lives. For example, what is your middle name? Most likely, the
moment you read this question, the name “popped” into your thoughts without any effort. But, in fact, there’s good reason to think that this simple bit of
remembering requires a complex series of steps. These steps take place outside
of awareness; and so, if we rely on introspection as our means of studying
mental events, we have no way of examining these processes.
The Cognitive Revolution
•
9
But there’s another, deeper problem with introspection. In order for any
science to proceed, there must be some way to test its claims; otherwise, we
have no means of separating correct assertions from false ones, accurate
descriptions of the world from fictions. Along with this requirement, science needs some way of resolving disagreements. If you claim that Earth
has one moon and I insist that it has two, we need some way of determining who is right. Otherwise, our “science” will become a matter of opinion,
not fact.
With introspection, this testability of claims is often unattainable. To see
why, imagine that I insist my headaches are worse than yours. How could
we ever test my claim? It might be true that I describe my headaches in
extreme terms: I talk about them being “agonizing” and “excruciating.” But
that might indicate only that I like to use extravagant descriptions; those
words might reveal my tendency to exaggerate (or to complain), not the
actual severity of my headaches. Similarly, it might be true that I need bed
rest whenever one of my headaches strikes. Does that mean my headaches are
truly intolerable? It might mean instead that I’m self-indulgent and rest even
when I feel mild pain. Perhaps our headaches are identical, but you’re stoic
about yours and I’m not.
How, therefore, should we test my claim about my headaches? What we
need is some way of directly comparing my headaches to yours, and that
would require transplanting one of my headaches into your experience,
or vice versa. Then one of us could make the appropriate comparison. But
(setting aside science fiction or fantasy) there’s no way to do this, leaving us,
in the end, unable to determine whether my headache reports are distorted or
accurate. We’re left, in other words, with the brute fact that our only information about my headaches is what comes through the filter of my description, and we have no way to know how (or whether) that filter is coloring
the evidence.
For purposes of science, this is unacceptable. Ultimately, we do want
to understand conscious experience, and so, in later chapters, we will consider introspective reports. For example, we’ll talk about the subjective
feeling of “familiarity” and the conscious experience of mental imagery; in
Chapter 14, we’ll talk about consciousness itself. In these settings, though,
we’ll rely on introspection as a source of observations that need to be
explained. We won’t rely on introspective data as a means of evaluating
our hypotheses — because, usually, we can’t. If we want to test hypotheses,
we need data we can rely on, and, among other requirements, this means
data that aren’t dependent on a particular point of view or a particular
descriptive style. Scientists generally achieve this objectivity by making
sure the raw data are out in plain view, so that you can inspect my evidence,
and I can inspect yours. In that way, we can be certain that neither of us is
distorting or misreporting the facts. And that is precisely what we cannot
do with introspection.
10 •
C H A P T E R O N E The Science of the Mind
The Years of Behaviorism
Historically, the concerns just described led many psychologists to abandon
introspection as a research tool. Psychology couldn’t be a science, they
argued, if it relied on this method. Instead, psychology needed objective data,
and that meant data out in the open for all to observe.
What sorts of data does this allow? First, an organism’s behaviors are
observable in the right way: You can watch my actions, and so can anyone
else who is appropriately positioned. Therefore, data concerned with behavior are objective data and thus grist for the scientific mill. Likewise, stimuli
in the world are in the same “objective” category: These are measurable,
recordable, physical events.
In addition, you can arrange to record the stimuli I experience day after
day after day and also the behaviors I produce each day. This means that
you can record how the pattern of my behavior changes over time and with
the accumulation of experience. In other words, my learning history can be
objectively recorded and scientifically studied.
In contrast, my beliefs, wishes, goals, preferences, hopes, and expectations
cannot be directly observed, cannot be objectively recorded. These “mentalistic” notions can be observed only via introspection; and introspection,
we’ve suggested, has little value as a scientific tool. Therefore, a scientific
psychology needs to avoid these invisible internal entities.
This perspective led to the behaviorist movement, a movement that dominated psychology in America for the first half of the 20th century. The movement
JOHN B. WATSON
John B. Watson (1878–1958) was a prominent
and persuasive advocate for the behaviorist movement. Given his focus on learning
and learning histories, it’s not surprising that
Watson was intrigued by babies’ behavior
and learning. Here, he tests the grasp reflex
displayed by human infants.
The Cognitive Revolution
•
11
was in many ways successful and uncovered a range of principles concerned
with how behavior changes in response to various stimuli (including the
stimuli we call “rewards” and “punishments”). By the late 1950s, however, psychologists were convinced that a lot of our behavior could not be explained
in these terms. The reason, basically, is that the ways people act, and the ways
they feel, are guided by how they understand or interpret the situation, and
not by the objective situation itself. Therefore, if we follow the behaviorists’
instruction and focus only on the objective situation, we will often misunderstand why people are doing what they’re doing and make the wrong predictions about how they’ll behave in the future. To put this point another way,
the behaviorist perspective demands that we not talk about mental entities
such as beliefs, memories, and so on, because these things cannot be studied
directly and so cannot be studied scientifically. Yet it seems that these subjective entities play a pivotal role in guiding behavior, and so we must consider
them if we want to understand behavior.
Evidence pertinent to these assertions is threaded throughout the chapters
of this book. Over and over, we’ll find it necessary to mention people’s perceptions and strategies and understanding, as we explain why (and how)
they perform various tasks and accomplish various goals. Indeed, we’ve
already seen an example of this pattern. Imagine that we present the “Betsy and
Jacob” story to people and then ask various questions: Why did Betsy shake
her piggy bank? Why did she go to look for her mother? People’s responses
will surely reflect their understanding of the story, which in turn depends on
far more than the physical stimulus — that is, the 29 syllables of the story
itself. If we want to predict someone’s responses to these questions, therefore,
we’ll need to refer to the stimulus (the story itself) and also to the person’s
knowledge and understanding of this stimulus.
Here’s a different example that makes the same general point. Imagine
you’re sitting in the dining hall. A friend produces this physical stimulus:
“Pass the salt, please,” and you immediately produce a bit of salt-passing
behavior. In this exchange, there is a physical stimulus (the words your friend
uttered) and an easily defined response (your passing of the salt), and so
this simple event seems fine from the behaviorists’ perspective — the elements
are out in the open, for all to observe, and can be objectively recorded. But
note that the event would have unfolded in the same way if your friend had
offered a different stimulus. “Could I have the salt?” would have done the
trick. Ditto for “Salt, please!” or “Hmm, this sure needs salt!” If your friend is
both loquacious and obnoxious, the utterance might have been: “Excuse me,
but after briefly contemplating the gustatory qualities of these comestibles, I
have discerned that their sensory qualities would be enhanced by the addition
of a number of sodium and chloride ions, delivered in roughly equal proportions and in crystalline form; could you aid me in this endeavor?” You might
giggle (or snarl) at your friend, but you would still pass the salt.
Now let’s work on the science of salt-passing behavior. When is this
behavior produced? We’ve just seen that the behavior is evoked by a number
of different stimuli, and so we would surely want to ask: What do these
12 •
C H A P T E R O N E The Science of the Mind
stimuli have in common? If we can answer that question, we’re on our way
to understanding why these stimuli all have the same effect.
The problem, though, is that if we focus on the observable, objective
aspects of these stimuli, they actually have little in common. After all, the
sounds being produced in that long statement about sodium and chloride
ions are rather different from the sounds in the utterance “Salt, please!” And
in many circumstances, similar sounds would not lead to salt-passing behavior. Imagine that your friend says, “Salt the pass” or “Sass the palt.” These
are acoustically similar to “Pass the salt” but wouldn’t have the same impact.
Or imagine that your friend says, “She has only a small part in the play. All
she gets to say is ‘Pass the salt, please.’” In this case, the right syllables were
uttered, but you wouldn’t pass the salt in response.
It seems, then, that our science of salt passing won’t get very far if we
insist on talking only about the physical stimulus. Stimuli that are physically
different from each other (“Salt, please” and the bit about ions) have similar
effects. Stimuli that are physically similar to each other (“Pass the salt” and
“Sass the palt”) have different effects. Physical similarity, therefore, is not
what unites the various stimuli that evoke salt passing.
It’s clear, though, that the various stimuli that evoke salt passing do have
something in common: They all mean the same thing. Sometimes this meaning derives from the words themselves (“Please pass the salt”). In other cases,
the meaning depends on certain pragmatic rules. (For example, you understand that the question “Could you pass the salt?” isn’t a question about arm
strength, although, interpreted literally, it might be understood that way.) In
all cases, though, it seems plain that to predict your behavior in the dining
hall, we need to ask what these stimuli mean to you. This seems an extraordinarily simple point, but it is a point, echoed by countless other examples, that
indicates the impossibility of a complete behaviorist psychology.1
PASSING THE SALT
If a friend requests the salt,
your response will depend
on how you understand your
friend’s words. This is a simple point, echoed in example
after example, but it is the
reason why a rigid behaviorist perspective cannot explain
your behavior.
The Intellectual Foundations of the
Cognitive Revolution
One might think, then, that we’re caught in a trap. On one side, it seems that the
way people act is shaped by how they perceive the situation, how they understand
the stimuli, and so on. If we want to explain behavior, then, we have no choice.
We need to talk about the mental world. But, on the other side, the only direct
means of studying the mental world is introspection, and introspection is scientifically unworkable. Therefore: We need to study the mental world, but we can’t.
There is, however, a solution to this impasse, and it was suggested years ago
by the philosopher Immanuel Kant (1724–1804). To use Kant’s transcendental
method, you begin with the observable facts and then work backward from
1. The behaviorists themselves quickly realized this point. As a result, modern behaviorism has
abandoned the radical rejection of mentalistic terms; indeed, it’s hard to draw a line between
modern behaviorism and a field called “animal cognition,” a field that often uses mentalistic
language! The behaviorism being criticized here is a historically defined behaviorism, and it’s
this perspective that, in large measure, gave birth to modern cognitive psychology.
The Cognitive Revolution
•
13
IMMANUEL KANT
Philosopher Immanuel Kant
(1724–1804) made major contributions to many fields,
and his transcendental method
enabled him to ask what
qualities of the mind make
experience possible.
14 •
these observations. In essence, you ask: How could these observations have
come about? What must be the underlying causes that led to these effects?
This method, sometimes called “inference to best explanation,” is at the
heart of most modern science. Physicists, for example, routinely use this
method to study objects or events that cannot be observed directly. To take just
one case, no physicist has ever observed an electron, but this hasn’t stopped
physicists from learning a great deal about electrons. How do the physicists
proceed? Even though electrons themselves aren’t observable, their presence
often leads to observable results — in essence, visible effects from an invisible
cause. For example, electrons leave observable tracks in cloud chambers, and
they can produce momentary fluctuations in a magnetic field. Physicists can
then use these observations in the same way a police detective uses clues —
asking what the “crime” must have been like if it left this and that clue.
(A size 11 footprint? That probably tells us what size feet the criminal has,
even though no one saw his feet. A smell of tobacco smoke? That suggests the
criminal was a smoker. And so on.) In the same way, physicists observe the clues
that electrons leave behind, and from this information they form hypotheses
about what electrons must be like in order to have produced those effects.
Of course, physicists (and other scientists) have a huge advantage over a
police detective. If the detective has insufficient evidence, she can’t arrange
for the crime to happen again in order to produce more evidence. (She can’t
say to the robber, “Please visit the bank again, but this time don’t wear
a mask.”) Scientists, in contrast, can arrange for a repeat of the “crime”
they’re seeking to explain — they can arrange for new experiments, with
new measures. Better still, they can set the stage in advance, to maximize
the likelihood that the “culprit” (in our example, the electron) will leave
useful clues behind. They can, for example, add new recording devices to the
situation, or they can place various obstacles in the electron’s path. In this
way, scientists can gather more and more data, including data crucial for
testing the predictions of a particular theory. This prospect — of reproducing
experiments and varying the experiments to test hypotheses — is what gives
science its power. It’s what enables scientists to assert that their hypotheses
have been rigorously tested, and it’s what gives scientists assurance that their
theories are correct.
Psychologists work in the same fashion — and the notion that we could
work in this fashion was one of the great contributions of the cognitive revolution. The idea is this: We know that we need to study mental processes;
that’s what we learned from the limitations of classical behaviorism. But we
also know that mental processes cannot be observed directly; we learned that
from the downfall of introspection. Our path forward, therefore, is to study
mental processes indirectly, relying on the fact that these processes, themselves invisible, have visible consequences: measurable delays in producing a
response, performances that can be assessed for accuracy, errors that can be
scrutinized and categorized. By examining these (and other) effects produced
by mental processes, we can develop — and test — hypotheses about what the
mental processes must have been. In this way, we use Kant’s method, just as
C H A P T E R O N E The Science of the Mind
physicists (or biologists or chemists or astronomers) do, to develop a science
that does not rest on direct observation.
The Path from Behaviorism to the
Cognitive Revolution
In setting after setting, cognitive psychologists have applied the Kantian logic
to explain how people remember, make decisions, pay attention, or solve
problems. In each case, we begin with a particular performance — say, a problem that someone solved — and then hypothesize a series of unseen mental
events that made the performance possible. But we don’t stop there. We also
ask whether some other, perhaps simpler, sequence of events might explain
the data. In other words, we do more than ask how the data came about; we
seek the best way to think about the data.
This pattern of theorizing has become the norm in psychology — a powerful indication that the cognitive revolution did indeed change the entire field.
But what triggered the revolution? What happened in the 1950s and 1960s
that propelled psychology forward in this way? It turns out that multiple
forces were in play.
One contribution came from within the behaviorist movement itself. We’ve
discussed concerns about classical behaviorism, and some of those concerns
were voiced early on by Edward Tolman (1886-1959) — a researcher who can
be counted both as a behaviorist and as one of the forerunners of cognitive
psychology. Prior to Tolman, most behaviorists argued that learning could
be understood simply as a change in behavior. Tolman argued, however, that
learning involved something more abstract: the acquisition of new knowledge.
In one of Tolman’s studies, rats were placed in a maze day after day. For
the initial 10 days, no food was available anywhere in the maze, and the rats
wandered around with no pattern to their behavior. Across these days, therefore, there was no change in behavior — and so, according to the conventional
view, no learning. But, in fact, there was learning, because the rats were learning the layout of the maze. That became clear on the 11th day of testing, when
food was introduced into the maze in a particular location. The next day, the
rats, placed back in the maze, ran immediately to that location. Indeed, their
behavior was essentially identical to the behavior of rats who had had many
days of training with food in the maze (Tolman, 1948; Gleitman, 1963).
What happened here? Across the initial 10 days, rats were acquiring what
Tolman called a “cognitive map” of the maze. In the early days of the procedure, however, the rats had no motivation to use this knowledge. On Days 11
and 12, though, the rats gained a reason to use what they knew, and at that
point they revealed their knowledge. The key point, though, is that — even for
rats — we need to talk about (invisible) mental processes (e.g., the formation
of cognitive maps) if we want to explain behavior.
A different spur to the cognitive revolution also arose out of behaviorism —
but this time from a strong critique of behaviorism. B.F. Skinner (1904–1990)
was an influential American behaviorist, and in 1957 he applied his style of
ULRIC NEISSER
Many intellectual developments led to the cognitive
revolution. A huge boost,
though, came from Ulric
Neisser’s book, Cognitive
Psychology (1967). Neisser’s
influence was so large that
many scholars refer to him
as the “father of cognitive
psychology.”
The Cognitive Revolution
•
15
analysis to humans’ ability to learn and use language, arguing that language
use could be understood in terms of behaviors and rewards (Skinner, 1957).
Two years later, the linguist Noam Chomsky (1928– ) published a ferocious
rebuttal to Skinner’s proposal, and convinced many psychologists that an
entirely different approach was needed for explaining language learning and
language use, and perhaps for other achievements as well.
European Roots of the Cognitive Revolution
Research psychology in the United States was, we’ve said, dominated by the
behaviorist movement for many years. The influence of behaviorism was not
as strong, however, in Europe, and several strands of European research fed
into and strengthened the cognitive revolution. In Chapter 3, we will describe
some of the theorizing that grew out of the Gestalt psychology movement, an
important movement based in Berlin in the early decades of the 20th century.
(Many of the Gestaltists fled to the United States in the years leading up to
World War II and became influential figures in their new home.) Overall, the
Gestalt psychologists argued that behaviors, ideas, and perceptions are organized in a way that could not be understood through a part-by-part, elementby-element, analysis of the world. Instead, they claimed, the elements take
on meaning only as part of the whole — and therefore psychology needed to
understand the nature of the “whole.” This position had many implications,
including an emphasis on the role of the perceiver in organizing his or her
experience. As we will see, this notion — that perceivers shape their own
experience — is a central theme for modern cognitive psychology.
Another crucial figure was British psychologist Frederic Bartlett (1886–1969).
Although he was working in a very different tradition from the Gestalt
psychologists, Bartlett also emphasized the ways in which each of us shapes
and organizes our experience. Bartlett claimed that people spontaneously fit
their experiences into a mental framework, or “schema,” and rely on this
schema both to interpret the experience as it happens and to aid memory
later on. We’ll say more about Bartlett’s work (found primarily in his book
Remembering, published in 1932) in Chapter 8.
FREDERIC BARTLETT
Frederic Bartlett was the
first professor of experimental psychology at the University of Cambridge. He is
best known for his studies
of memory and the notion
that people spontaneously
fit their experiences into a
“schema,” and they rely on
the schema both to guide
their understanding and (later)
to guide their memory.
16 •
Computers and the Cognitive Revolution
Tolman, Chomsky, the Gestaltists, and Bartlett disagreed on many points.
Even so, a common theme ran through their theorizing: These scholars all
agreed that we could not explain humans’ (or even rats’) behavior unless we
explain what is going on within the mind — whether our emphasis is on cognitive maps, schemata, or some other form of knowledge. But, in explaining
this knowledge and how the knowledge is put to use, where should we begin?
What sorts of processes or mechanisms might we propose?
Here we meet another crucial stream that fed into the cognitive revolution,
because in the 1950s a new approach to psychological explanation became
available and turned out to be immensely fruitful. This new approach was
C H A P T E R O N E The Science of the Mind
suggested by the rapid developments in electronic information processing,
including developments in computer technology. It soon became clear that computers were capable of immensely efficient information storage and retrieval
(“memory”), as well as performance that seemed to involve decision making
and problem solving. Indeed, some computer scientists proposed that computers would soon be genuinely intelligent — and the field of “artificial intelligence”
was launched and made rapid progress (e.g., Newell & Simon, 1959).
Psychologists were intrigued by these proposals and began to explore the
possibility that the human mind followed processes and procedures similar
to those used in computers. As a result, psychological data were soon being
explained in terms of “buffers” and “gates” and “central processors,” terms
borrowed from computer technology (e.g., Miller, 1956; Miller, Galanter,
& Pribram, 1960). This approach was evident, for example, in the work of
another British psychologist, Donald Broadbent (1926–1993). He was one of
the earliest researchers to use the language of computer science in explaining human cognition. His work emphasized a succession of practical issues,
including the mechanisms through which people focus their attention when
working in complex environments, and his book Perception and Communication (1958) framed discussions of attention for many years.
This computer-based vocabulary allowed a new style of theorizing. Given
a particular performance, say, in paying attention or on some memory task,
one could hypothesize a series of information-processing events that made
the performance possible. As we will see, hypotheses cast in these terms led
psychologists to predict a broad range of new observations, and in this way
both organized the available information and led to many new discoveries.
TEST YOURSELF
3. W
hy is introspection
limited as a source of
scientific evidence?
4. W
hy do modern
psychologists agree
that we have to refer
to mental states (what
you believe, what you
perceive, what you
understand) in order
to explain behavior?
5. D
escribe at least one
historical development
that laid the groundwork for the cognitive
revolution.
Research in Cognitive Psychology:
The Diversity of Methods
Over the last half-century, cognitive psychologists have continued to frame
many hypotheses in these computer-based terms. But we’ve also developed
other options for theorizing. For example, before we’re done in this book,
we’ll also discuss hypotheses framed in terms of the strategies a person is
relying on, or the inferences she is making. No matter what the form of the
hypothesis, though, the next steps are crucial. First, we derive new predictions from the hypothesis, along the lines of “If this is the mechanism behind
the original findings, then things should work differently in this circumstance
or that one.” Then, we gather new data to test those predictions. If the data fit
with the predictions, this outcome confirms the hypothesis. If the data don’t
line up with the predictions, a new hypothesis is needed.
But what methods do we use, and what sorts of data do we collect? The
answer, in brief, is that we use diverse methods and collect many types of
data. In other words, what unites cognitive psychology is not an allegiance
to any particular procedure in the laboratory. Instead, what unites the
field is the logic that underlies our research, no matter what method we use.
Research in Cognitive Psychology: The Diversity of Methods
•
17
(We discuss this logic more fully in the appendix for this textbook. The
appendix contains a series of modules, with each module exploring an aspect
of research methodology directly related to one of the book’s chapters.)
What sorts of data do we use? In some settings, we ask how well people
perform a particular task. For example, in tests of memory we might ask
how complete someone’s memory is (does the person remember all of the
objects in view in a picture?) and also how accurate the memory is (does the
person perhaps remember seeing a banana when, in truth, no banana was in
view?). We can also ask how performance changes if we change the “input”
(how well does the person remember a story, rather than a picture?), and we
can change the person’s circumstances (how is memory changed if the person
is happy, or afraid, when hearing the story?). We can also manipulate the
person’s plans or strategies (what happens if we teach the person some sort
of memorization technique?), and we can compare different people (children
vs. adults; novices at a task vs. experts; people with normal vision vs. people
who have been blind since birth).
A different approach relies on measurements of speed. The idea here is
that mental operations are fast but do take a measurable amount of time, and
by examining the response time (RT) — that is, how long someone needs to
make a particular response — we can often gain important insights into what’s
going on in the mind. For example, imagine that we ask you: “Yes or no: Do
cats have whiskers?” And then: “Yes or no: Do cats have heads?” Both questions are absurdly easy, so there’s no point in asking whether you’re accurate
in your responses — it’s a sure bet that you will be. We can, however, measure
your response times to questions like these, often with intriguing results. For
example, if you’re forming a mental picture of a cat when you’re asked these
questions, you’ll be faster for the “heads” question than the “whiskers” question. If you think about cats without forming a mental picture, the pattern
reverses — you’ll be faster for the “whiskers” question. In Chapter 11, we’ll
use results like these to test hypotheses about how information — and mental
pictures in particular — are represented and analyzed in your mind.
We can also gain insights from observations focused on the brain and
nervous system. Over the last few decades, cognitive psychology has formed
a productive partnership with the field of cognitive neuroscience, the effort
toward understanding humans’ mental functioning through close study of
the brain and nervous system. But here, too, numerous forms of evidence are
available. We’ll say more about these points in the next chapter, but for now
let’s note that we can learn a lot by studying people with damaged brains and
also people with healthy brains. Information about damaged brains comes
from the field of clinical neuropsychology, the study of brain function that
uses, as its main data source, cases in which damage or illness has disrupted
the working of some brain structure. We’ve already mentioned H.M., a man
whose memory was massively disrupted as an unexpected consequence of
surgery. As a different example, in Chapter 12 we’ll consider cases in which
someone’s ability to make ordinary decisions (Coke or Pepsi? Wear the blue
sweater or the green one?) is disrupted if brain centers involved in emotion
18 •
C H A P T E R O N E The Science of the Mind
are disrupted; observations like these provide crucial information about the
role of emotion in decision making.
Information about healthy brains comes from neuroimaging techniques,
which enable us, with some methods, to scrutinize the precise structure of the
brain and, with other methods, to track the moment-by-moment pattern of
activation within someone’s brain. We’ll see in Chapter 7, for example, that
different patterns of brain activation during learning lead to different types
of memory, and we’ll use this fact as we ask what the types of memory are.
There’s no reason for you, as a reader, to memorize this catalogue of different types of evidence. That’s because we’ll encounter each of these forms
of data again and again in this text. Our point for now is simply to highlight
the fact that there are multiple tools with which we can test, and eventually
confirm, various claims. Indeed, relying on these tools, cognitive psychology
has learned a tremendous amount about the mind. Our research has brought
us powerful new theories and enormously useful results. Let’s dive in and
start exploring what the science of the mind has taught us.
TEST YOURSELF
6. D
escribe at least three
types of evidence that
cognitive psychologists
routinely rely on.
APPLYING COGNITIVE PSYCHOLOGY
Research in cognitive psychology can help us understand deep theoretical
issues, such as what it means to be rational or what the function of consciousness might be. But our research also has broad practical implications,
and so our studies often provide lessons for how we should conduct our
daily lives.
Some of the practical lessons from cognitive psychology are obvious. For
example, research on memory can help students who are trying to learn new
materials in the classroom; studies of how people draw conclusions can help
people to draw smarter, more defensible conclusions. Following these leads,
each chapter in this text ends with an essay that explores how the material in
that chapter can be applied to an issue that’s important for education. This
emphasis is rooted, in part, in the fact that most readers of this book will be
college students, using the book in the context of one of their courses. I hope,
therefore, that the Cognitive Psychology and Education essays are directly
useful for these readers! Concretely, the essay at the end of Chapter 4, for
example, will teach you how to speed-read (but will also explain the limitations of speed-reading). The essay at the end of Chapter 6 will offer suggestions for how to study and retain the material you’re hoping to learn.
Let’s emphasize, though, that research in cognitive psychology also has
implications for other domains. For example, think about the criminal justice
system and what happens in a criminal investigation. Eyewitnesses provide
evidence, based on what they paid attention to during a crime and what they
remember. Police officers question the witnesses, trying to get the most out
of what each witness recalls — but without leading the witness in any way.
Then, the police try to deduce, from the evidence, who the perpetrator was.
Applying Cognitive Psychology
•
19
Later, during the trial, jurors listen to evidence and make a judgment about
the defendant’s innocence or guilt.
Cast in these terms, it should be obvious that an understanding of
attention, memory, reasoning, and judgment (to name just a few processes) is
directly relevant to what happens in the legal system. On this basis, therefore,
I’ve also written essays that focus on the interplay between cognitive psychology and the law. The essay for Chapter 3, for example, uses what we know
about visual perception to ask what we can expect witnesses to see. The essay
for Chapter 7 explores a research-based procedure for helping witnesses to
recall more of what they’ve observed. If you’re curious to see these Cognitive
Psychology and the Law essays, you can find them online in the ebook, available at http://digital.wwnorton.com/cognition7.
COGNITIVE PSYCHOLOGY AND THE CRIMINAL JUSTICE SYSTEM
Eyewitnesses in the courtroom rely on what they remember about key events, and
what they remember depends crucially on what they perceived and paid attention
to. Therefore, our understanding of memory, perception, and attention can help the
justice system in its evaluation of witness evidence.
20 •
C H A P T E R O N E The Science of the Mind
chapter review
SUMMARY
• Cognitive psychology is concerned with how
people remember, pay attention, and think. The
importance of all these issues arises partly from
the fact that most of what we do, say, and feel is
guided by things we already know. One example
is our comprehension of a simple story, which turns
out to be heavily influenced by the knowledge
we supply.
• Cognitive psychology emerged as a separate discipline in the late 1950s, and its powerful impact
on the wider field of psychology has led many academics to speak of this emergence as the cognitive
revolution. One predecessor of cognitive psychology was the 19th-century movement that emphasized introspection as the main research tool for
psychology. But psychologists soon became disenchanted with this movement for several reasons:
Introspection cannot inform us about unconscious
mental events; and even with conscious events,
claims rooted in introspection are often untestable because there is no way for an independent
observer to check the accuracy or completeness of
an introspective report.
• The behaviorist movement rejected introspection
as a method, insisting instead that psychology speak
only of mechanisms and processes that are objective and out in the open for all to observe. However, evidence suggests that our thinking, behavior,
and feelings are often shaped by our perception or
understanding of the events we experience. This is
problematic for the behaviorists: Perception and
understanding are exactly the sorts of mental processes that the behaviorists regarded as subjective
and not open to scientific study.
• In order to study mental events, psychologists
have turned to a method in which one focuses on
observable events but then asks what (invisible)
events must have taken place in order to make these
(visible) effects possible.
• Many factors contributed to the emergence
of cognitive psychology in the 1950s and 1960s.
Tolman’s research demonstrated that even in rats,
learning involved the acquisition of new knowledge
and not just a change in behavior. Chomsky argued
powerfully that a behaviorist analysis was inadequate as an explanation for language learning and
language use. Gestalt psychologists emphasized
the role of the perceiver in organizing his or her
experience. Bartlett’s research showed that people
spontaneously fit their experiences into a mental
framework, or schema.
• Early theorizing in cognitive psychology often
borrowed ideas from computer science, including
early work on artificial intelligence.
• Cognitive psychologists rely on a diverse set of
methods and collect many types of data. Included
are measures of the quality of someone’s performance, measures of response speed, and, in some
cases, methods that allow us to probe the under­
lying biology.
21
KEY TERMS
introspection (p. 9)
behaviorist theory (p. 11)
transcendental method (p. 13)
response time (RT) (p. 18)
cognitive neuroscience (p. 18)
clinical neuropsychology (p. 18)
neuroimaging techniques (p. 19)
TEST YOURSELF AGAIN
1.Why is memory crucial for behaviors and
mental operations that don’t in any direct or
explicit way ask you “to remember”?
5.Describe at least one historical development
that laid the groundwork for the cognitive
revolution.
2.What aspects of H.M.’s life were disrupted as a
result of his amnesia?
6.Describe at least three types of evidence that
cognitive psychologists routinely rely on.
3.Why is introspection limited as a source of
scientific evidence?
4.Why do modern psychologists agree that we
have to refer to mental states (what you believe,
what you perceive, what you understand) in
order to explain behavior?
THINK ABOUT IT
1.The chapter argues that in a wide range of settings, our behaviors and our emotions depend
on what we know, believe, and remember. Can
you come up with examples of your own that
illustrate this reliance on cognition in a circumstance that doesn’t seem, on the surface, to be
one that involves “intellectual activity”?
22
2.Some critics of Darwin’s theory of evolution
via natural selection argue this way: “Darwin’s
claims can never be tested, because of course no
one was around to observe directly the processes
of evolution that Darwin proposed.” Why
is this assertion misguided, resting on a false
notion of how science proceeds?
E eBOOK DEMONSTRATIONS & ESSAYS
Go to http://digital.wwnorton.com/cognition7 for the online demonstrations and essays
relevant to this chapter. These can be found in the ebook.
Online Applying Cognitive Psychology and the
Law Essays
• Cognitive Psychology and the Law: Improving
the Criminal Justice System
COGNITION LABS
Go to http://digital.wwnorton.com/cognition7 to experience interactive online psychology
labs relevant to this chapter.
23
2
chapter
The Neural Basis
for Cognition
what if…
Throughout this text, we’ll be examining ordinary
achievements. A friend asks: “Where’d you grow up?”
and you immediately answer. You’re meeting a friend at the airport,
and you instantly recognize her the moment she steps into view. An
instructor says, “Listen carefully,” and you have no trouble focusing
your attention.
Ordinary or not, achievements like these are crucial for you, and
your life would be massively disrupted if you couldn’t draw information from memory, or recognize the objects you encounter, or choose
where you’ll point your attention. As a way of dramatizing this point,
we’ll begin each chapter by asking: What would happen to someone if
one of these fundamental capacities didn’t work as it normally does?
What if . . . ?
The disorder known as Capgras syndrome (Capgras & ReboulLachaux, 1923) is relatively rare, but it can result from various injuries
to the brain (Ellis & De Pauw, 1994) and is sometimes found in people
with Alzheimer’s syndrome (Harwood, Barker, Ownby, & Duara, 1999).
Someone with this syndrome is fully able to recognize the people in her
world — her husband, her parents, her friends — but is utterly convinced
that these people are not who they appear to be. The real husband or
the real son, the afflicted person insists, has been kidnapped (or worse).
The person now in view, therefore, must be a fraud of some sort, impersonating the (allegedly) absent person.
Imagine what it’s like to have this disorder. You turn to your father
and exclaim, “You look like my father, sound like him, and act like him.
But I can tell that you’re not my father! Who are you?”
Often, a person with Capgras syndrome insists that there are slight
differences between the “impostor” and the person he (or she) has
supposedly replaced — subtle changes in personality or appearance.
Of course, no one else detects these (nonexistent) differences, which
can lead to paranoid suspicions about why a loved one has been
taken away and why no one else will acknowledge the replacement.
In the extreme, these suspicions can lead a Capgras sufferer to desperate steps. In some cases, patients suffering from this syndrome
have murdered the supposed impostor in an attempt to end the
charade and relocate the “genuine” character. In one case, a Capgras
patient was convinced his father had been replaced by a robot and so
25
preview of chapter themes
•
e begin by exploring the example of Capgras syndrome
W
to illustrate how seemingly simple achievements actually
depend on many parts of the brain. We also highlight the
ways that the study of the brain can illuminate questions
about the mind.
•
e then survey the brain’s anatomy, emphasizing the
W
function carried out by each region. Identification of these
functions is supported by neuroimaging data, which can
assess the activity levels in different areas, and by studies
of the effects of brain damage.
•
e then take a closer look at the various parts of the cereW
bral cortex — the most important part of the brain for cognitive functioning. These parts include the motor areas,
the sensory areas, and the so-called association cortex.
•
inally, we turn to the individual cells that make up the
F
brain — the neurons and glia — and discuss the basic principles of how these cells function.
decapitated him in order to look for the batteries and microfilm in his
head (Blount, 1986).
What is going on here? The answer lies in the fact that facial recognition involves two separate systems in the brain. One system leads to
a cognitive appraisal (“I know what my father looks like, and I can perceive that you closely resemble him”), and the other to a more global,
emotional appraisal (“You look familiar to me and also trigger a warm
response in me”). When these two appraisals agree, the result is a confident recognition (“You obviously are my father”). In Capgras syndrome,
though, the emotional processing is disrupted, leading to an intellectual
identification without a familiarity response (Ellis & Lewis, 2001; Ellis &
Young, 1990; Ramachandran & Blakeslee, 1998): “You resemble my father
but trigger no sense of familiarity, so you must be someone else.” The
result? Confusion and, at times, bizarre speculation about why a loved
one has been kidnapped and replaced — and a level of paranoia that can,
as we have seen, lead to homicide.
Explaining Capgras Syndrome
We began this chapter with a description of Capgras syndrome, and we’ve
offered an account of the mental processes that characterize this disorder.
Specifically, we’ve suggested that someone with this syndrome is able to recognize a loved one’s face, but with no feeling of familiarity. Is this the right
way to think about Capgras syndrome?
One line of evidence comes from neuroimaging techniques that enable
researchers to take high-quality, three-dimensional “pictures” of living brains
without in any way disturbing the brains’ owners. We’ll have more to say
about neuroimaging later; but first, what do these techniques tell us about
Capgras syndrome?
26 •
C H A P T E R T WO The Neural Basis for Cognition
The Neural Basis for Capgras Syndrome
Some types of neuroimaging provide portraits of the physical makeup of
the brain: What’s where? How are structures shaped or connected to each
other? Are there structures present (such as tumors) that shouldn’t be there,
or structures that are missing (because of disease or birth defects)? This information about structure was gained in older studies from positron emission tomography (more commonly referred to as a PET scan). More recent
studies usually rely on magnetic resonance imaging (MRI; see Figure 2.1).
These scans suggest a link between Capgras syndrome and abnormalities in
several brain areas, indicating that our account of the syndrome will need to
consider several elements (Edelstyn & Oyebode, 1999; also see O’Connor,
Walbridge, Sandson, & Alexander, 1996).
FIGURE 2.1
NEUROIMAGING
Scanners like this one are used for both MRI and fMRI scans. MRI scans tell us
about the structure of the brain; fMRI scans tell us which portions of the brain
are especially active during the scan. An fMRI scan usually results in color
images, with each hue indicating a particular activity level.
Explaining Capgras Syndrome
•
27
FIGURE 2.2
THE LOBES OF THE HUMAN BRAIN
Central fissure
Parietal lobe
Frontal lobe
Occipital
lobe
Lateral fissure
Temporal lobe
Cerebellum
A
B
Panel A identifies the various lobes and some of the brain’s prominent features. Actual brains, however, are uniformly colored, as shown in the photograph in Panel B. The four lobes of the forebrain surround (and hide from
view) the midbrain and most of the hindbrain. (The cerebellum is the only part of the hindbrain that is visible in the
figure, and, in fact, the temporal lobe has been pushed upward a bit in the left panel to make the cerebellum more
visible.) This side view shows the left cerebral hemisphere; the structures on the right side of the brain are similar.
However the two halves of the brain have somewhat different functions, and so the results of brain injury depend
on which half is damaged. The symptoms of Capgras syndrome, for example, result from damage to specific sites
on the right side of the frontal and temporal lobes.
One site of damage in Capgras patients is in the temporal lobe (see
Figure 2.2), particularly on the right side of the head. This damage probably disrupts circuits involving the amygdala, an almond-shaped structure
that — in the intact brain — seems to serve as an “emotional evaluator,”
helping an organism detect stimuli associated with threat or danger (see
Figure 2.3). The amygdala is also important for detecting positive stimuli —
indicators of safety or of available rewards. With damaged amygdalae,
therefore, people with Capgras syndrome won’t experience the warm sense
of feeling good (and safe and secure) when looking at a loved one’s familiar
28 •
C H A P T E R T WO The Neural Basis for Cognition
FIGURE 2.3 THE
AMYGDALA AS AN
“EMOTIONAL EVALUATOR”
The area shown in yellow marks
the location of the amygdala. In
this image, the yellow is a reflection of increased activity created
by a fear memory — the memory
of receiving an electric shock.
face. This lack of an emotional response is probably why these faces don’t
feel familiar to them, and is fully in line with the two-systems hypothesis
we’ve already sketched.
Patients with Capgras syndrome also have brain abnormalities in the
frontal lobe, specifically in the right prefrontal cortex. What is this area’s
normal function? To find out, we turn to a different neuroimaging technique,
functional magnetic resonance imaging (fMRI), which enables us to track
moment-by-moment activity levels in different sites in a living brain. (We’ll
say more about fMRI in a later section.) This technique allows us to answer
such questions as: When a person is reading, which brain regions are particularly active? How about when a person is listening to music? With data like
these, we can ask which tasks make heavy use of a brain area, and from that
base we can draw conclusions about that brain area’s function.
Studies make it clear that the prefrontal cortex is especially active when
a person is doing tasks that require planning or careful analysis. Conversely,
this area is less active when someone is dreaming. Plausibly, this latter pattern
Explaining Capgras Syndrome
•
29
reflects the absence of careful analysis of the dream material, which helps
explain why dreams are often illogical or bizarre.
Related, consider fMRI scans of patients suffering from schizophrenia
(e.g., Silbersweig et al., 1995). Neuroimaging reveals diminished activity in
the frontal lobes whenever these patients are experiencing hallucinations. One
interpretation is that the diminished activity reflects a decreased ability to distinguish internal events (thoughts) from external ones (voices) or to distinguish
imagined events from real ones (cf. Glisky, Polster, & Routhieaux, 1995).
How is all of this relevant to Capgras syndrome? With damage to the
frontal lobe, Capgras patients may be less able to keep track of what is real
and what is not, what is sensible and what is not. As a result, weird beliefs
can emerge unchecked, including delusions (about robots and the like) that
you or I would find totally bizarre.
What Do We Learn from Capgras Syndrome?
Other lines of evidence add to our understanding of Capgras syndrome
(e.g., Ellis & Lewis, 2001; Ramachandran & Blakeslee, 1998). Some of the
evidence comes from the psychology laboratory and confirms the suggestion that recognition of all stimuli (not just faces) involves two separate
mechanisms — one that hinges on factual knowledge, and one that’s more
“emotional” and tied to the warm sense of familiarity (see Chapter 7).
Note, then, that our understanding of Capgras syndrome depends on a
combination of evidence drawn from cognitive psychology and from cognitive neuroscience. We use both perspectives to test (and, ultimately, to confirm) the hypothesis we’ve offered. In addition, just as both perspectives can
illuminate Capgras syndrome, both can be illuminated by the syndrome. That
is, we can use Capgras syndrome (and other biological evidence) to illuminate broader issues about the nature of the brain and of the mind.
For example, Capgras syndrome suggests that the amygdala plays a crucial
role in supporting the feeling of familiarity. Other evidence suggests that the
amygdala also helps people remember the emotional events of their lives (e.g.,
Buchanan & Adolphs, 2004). Still other evidence indicates that the amygdala plays a role in decision making (e.g., Bechara, Damasio, & Damasio,
2003), especially for decisions that rest on emotional evaluations of one’s
options. Facts like these tell us a lot about the various functions that make
cognition possible and, more specifically, tell us that our theorizing needs to
include a broadly useful “emotional evaluator,” involved in many cognitive
processes. Moreover, Capgras syndrome tells us that this emotional evaluator works in a fashion separate from the evaluation of factual information,
and this observation gives us a way to think about occasions in which your
evaluation of the facts points toward one conclusion, while an emotional
evaluation points toward a different conclusion. These are valuable clues
as we try to understand the processes that support ordinary remembering
or decision making. (For more on the role of emotion in decision making,
see Chapter 12.)
30 •
C H A P T E R T WO The Neural Basis for Cognition
What does Capgras syndrome teach us about the brain itself? One lesson
involves the fact that many different parts of the brain are needed for even
the simplest achievement. In order to recognize your father, for example, one
part of your brain needs to store the factual memory of what he looks like.
Another part of the brain is responsible for analyzing the visual input you
receive when looking at a face. Yet another brain area has the job of comparing this now-analyzed input to the factual information provided from
memory, to determine whether there’s a match. Another site provides the
emotional evaluation of the input. A different site presumably assembles the
data from all these other sites — and registers the fact that the face being
inspected does match the factual recollection of your father’s face, and also
produces a warm sense of familiarity.
Usually, all these brain areas work together, allowing the recognition of
your father’s face to go smoothly forward. If they don’t work together — that
is, if coordination among these areas is disrupted — yet another area works
to make sure you offer reasonable hypotheses about this disconnect, and not
zany ones. (In other words, if your father looks less familiar to you on some
occasion, you’re likely to explain this by saying, “I guess he must have gotten
new glasses” rather than “I bet he’s been replaced by a robot.”)
Unmistakably, this apparently easy task — seeing your father and recognizing
who he is — requires multiple brain areas. The same is true of most tasks, and
in this way Capgras syndrome illustrates this crucial aspect of brain function.
TEST YOURSELF
1. W
hat are the symptoms of Capgras
syndrome, and why
do they suggest a
two-part explanation
for how you recognize
faces?
The Study of the Brain
In order to discuss Capgras syndrome, we needed to refer to different brain
areas and had to rely on several different research techniques. In this way, the
syndrome also illustrates another point — that this is a domain in which we
need some technical foundations before we can develop our theories. Let’s
start building those foundations.
The human brain weighs (on average) a bit more than 3 pounds (roughly
1.4 kg), with male brains weighing about 10% more than female brains
(Hartmann, Ramseier, Gudat, Mihatsch, & Polasek, 1994). The brain is roughly
the size of a small melon, yet this compact structure has been estimated to
contain 86 billion nerve cells (Azevedo et al., 2009). Each of these cells is connected to 10,000 or so others — for a total of roughly 860 trillion connections.
The brain also contains a huge number of glial cells, and we’ll have more to say
about all of these individual cells later on in the chapter. For now, though, how
should we begin our study of this densely packed, incredibly complex organ?
One place to start is with a simple fact we’ve already met: that different
parts of the brain perform different jobs. Scientists have known this fact
about the brain for many years, thanks to clinical evidence showing that
the symptoms produced by brain damage depend heavily on the location
of the damage. In 1848, for example, a horrible construction accident
caused Phineas Gage to suffer damage in the frontmost part of his brain
The Study of the Brain
•
31
FIGURE 2.4
A
PHINEAS GAGE
B
C
Phineas Gage was working as a construction foreman when some blasting powder misfired and launched a piece
of iron into his cheek and through the front part of his brain. Remarkably, Gage survived and continued to live a
fairly normal life, but his pattern of intellectual and emotional impairments provide valuable cues about the function of the brain’s frontal lobes. Panel A is a photo of Gage’s skull; the drawing in Panel B depicts the iron bar’s
path as it blasted through his head. Panel C is an actual photograph of Gage, and he’s holding the bar that went
through his brain!
(see Figure 2.4), and this damage led to severe personality and emotional
problems. In 1861, physician Paul Broca noted that damage in a different
location, on the left side of the brain, led to a disruption of language skills.
In 1911, Édouard Claparède (1911/1951) reported his observations with
patients who suffered from profound memory loss produced by damage in
still another part of the brain.
Clearly, therefore, we need to understand brain functioning with reference
to brain anatomy. Where was the damage that Gage suffered? Where was
the damage in Broca’s patients or Claparède’s? In this section, we fill in some
basics of brain anatomy.
Hindbrain, Midbrain, Forebrain
The human brain is divided into three main structures: the hindbrain,
the midbrain, and the forebrain. The hindbrain is located at the very top
of the spinal cord and includes structures crucial for controlling key life
32 •
C H A P T E R T WO The Neural Basis for Cognition
Corpus callosum
Midbrain
Pons
Medulla
Spinal cord
Cerebellum
GROSS ANATOMY OF A BRAIN SHOWING BRAIN STEM
The pons and medulla are part of the hindbrain. The medulla controls vital functions
such as breathing and heart rate. The pons (Latin for “bridge”) is the main connection between the cerebellum and the rest of the brain.
functions. It’s here, for example, that the rhythm of heartbeats and the
rhythm of breathing are regulated. The hindbrain also plays an essential
role in maintaining the body’s overall tone. Specifically, the hindbrain helps
maintain the body’s posture and balance; it also helps control the brain’s level
of alertness.
The largest area of the hindbrain is the cerebellum. For many years,
investigators believed this structure’s main role was in the coordination
of bodily movements and balance. Research indicates, however, that
the cerebellum plays various other roles and that damage to this organ
can cause problems in spatial reasoning, in discriminating sounds, and
in integrating the input received from various sensory systems (Bower &
Parsons, 2003).
The midbrain has several functions. It plays an important part in coordinating movements, including the precise movements of the eyes as they
explore the visual world. Also in the midbrain are circuits that relay auditory
information from the ears to the areas in the forebrain where this information is processed and interpreted. Still other structures in the midbrain help
to regulate the experience of pain.
The Study of the Brain
•
33
For our purposes, though, the most interesting brain region (and, in
humans, the largest region) is the forebrain. Drawings of the brain (like the
one shown in Figure 2.2) show little other than the forebrain, because this
structure surrounds (and so hides from view) the entire midbrain and most
of the hindbrain. Of course, only the outer surface of the forebrain — the
cortex — is visible in such pictures. In general, the word “cortex” (from
the Latin word for “tree bark”) refers to an organ’s outer surface, and many
organs each have their own cortex; what’s visible in the drawing, then, is the
cerebral cortex.
The cortex is just a thin covering on the outer surface of the forebrain; on average, it’s a mere 3 mm thick. Nonetheless, there’s a great
deal of cortical tissue; by some estimates, the cortex makes up 80% of the
human brain. This considerable volume is made possible by the fact
that the cerebral cortex, thin as it is, consists of a large sheet of tissue. If
stretched out flat, it would cover more than 300 square inches, or roughly
2,000 cm2. (For comparison, this is an area roughly 20% greater than
the area covered by an extra-large — 18 inch, or 46 cm — pizza.) But the
cortex isn’t stretched flat; instead, it’s crumpled up and jammed into the
limited space inside the skull. It’s this crumpling that produces the brain’s
most obvious visual feature — the wrinkles, or convolutions, that cover the
brain’s outer surface.
Some of the “valleys” between the wrinkles are actually deep grooves that
divide the brain into different sections. The deepest groove is the longitudinal
fissure, running from the front of the brain to the back, which separates the
left cerebral hemisphere from the right. Other fissures divide the cortex in
each hemisphere into four lobes (again, look back at Figure 2.2), and these
are named after the bones that cover them — bones that, as a group, make
up the skull. The frontal lobes form the front of the brain, right behind the
forehead. The central fissure divides the frontal lobes on each side of the
brain from the parietal lobes, the brain’s topmost part. The bottom edge of
the frontal lobes is marked by the lateral fissure, and below it are the temporal lobes. Finally, at the very back of the brain, connected to the parietal and
temporal lobes, are the occipital lobes.
Subcortical Structures
Hidden from view, underneath the cortex, are several subcortical structures.
One of these structures, the thalamus, acts as a relay station for nearly all
the sensory information going to the cortex. Directly underneath the thalamus is the hypothalamus, a structure that plays a crucial role in controlling
behaviors that serve specific biological needs — behaviors that include eating,
drinking, and sexual activity.
Surrounding the thalamus and hypothalamus is another set of structures
that form the limbic system. Included here is the amygdala, and close by
is the hippocampus, both located underneath the cortex in the temporal
lobe (plurals: amygdalae and hippocampi; see Figure 2.5). These structures
34 •
C H A P T E R T WO The Neural Basis for Cognition
FIGURE 2.5
T HE LIMBIC SYSTEM AND THE
HIPPOCAMPUS
Cingulate cortex
Fornix
Thalamus
Mamillary
body
Hypothalamus
Amygdala
Hippocampus
Color is used in this drawing to help you visualize the arrangement of these
brain structures. Imagine that the cortex is semitransparent, allowing you to
look into the brain to see the (subcortical) structures highlighted here. The
limbic system includes a number of subcortical structures that play a crucial
role in learning and memory and in emotional processing.
are essential for learning and memory, and the patient H.M., discussed in
Chapter 1, developed his profound amnesia after surgeons removed large
portions of these structures — strong confirmation of their role in the formation of new memories.
We mentioned earlier that the amygdala plays a key role in emotional
processing, and this role is reflected in many findings. For example, presentation of frightful faces causes high levels of activity in the amygdala
(Williams et al., 2006). Likewise, people ordinarily show more complete,
longer-lasting memories for emotional events, compared to similar but
emotionally flat events. This memory advantage for emotional events is
especially pronounced in people who showed greater activation in the
amygdala while they were witnessing the event in the first place. Conversely,
the memory advantage for emotional events is diminished (and may not be
observed at all) in people who (through sickness or injury) have suffered
damage to the amygdalae.
The Study of the Brain
•
35
Lateralization
TEST YOURSELF
2.What is the cerebral
cortex?
3.What are the four
major lobes of the
forebrain?
4.Identify some of the
functions of the hippocampus, the amygdala, and the corpus
callosum.
36 •
Virtually all parts of the brain come in pairs, and so there is a hippocampus
on the left side of the brain and another on the right, a left-side amygdala
and a right-side one. The same is true for the cerebral cortex itself: There
is a temporal cortex (i.e., a cortex of the temporal lobe) in the left hemisphere and another in the right, a left occipital cortex and a right one, and
so on. In all cases, cortical and subcortical, the left and right structures
in each pair have roughly the same shape and the same pattern of connections to other brain areas. Even so, there are differences in function
between the left-side and right-side structures, with each left-hemisphere
structure playing a somewhat different role from the corresponding righthemisphere structure.
Let’s remember, though, that the two halves of the brain work together —
the functioning of one side is closely integrated with that of the other side.
This integration is made possible by the commissures, thick bundles of
fibers that carry information back and forth between the two hemispheres.
The largest commissure is the corpus callosum, but several other structures
also make sure that the two brain halves work as partners in almost all
mental tasks.
In certain cases, though, there are medical reasons to sever the corpus
callosum and some of the other commissures. (For many years, this surgery was a last resort for extreme cases of epilepsy.) The person is then
said to be a “split-brain patient” — still having both brain halves, but with
communication between the halves severely limited. Research with these
patients has taught us a great deal about the specialized function of the
brain’s two hemispheres. It has provided evidence, for example, that many
aspects of language processing are lodged in the left hemisphere, while
the right hemisphere seems crucial for a number of tasks involving spatial
judgment (see Figure 2.6).
However, it’s important not to overstate the contrast between the two
brain halves, and it’s misleading to claim (as some people do) that we need
to silence our “left-brain thinking” in order to be more creative, or that
intuitions grow out of “right-brain thinking.” These claims do begin with a
kernel of truth, because some elements of creativity depend on specialized
processing in the right hemisphere (see, e.g., Kounios & Beeman, 2015).
Even so, whether we’re examining creativity or any other capacity, the two
halves of the brain have to work together, with each hemisphere making its
own contribution to the overall performance. Therefore, “shutting down” or
“silencing” one hemisphere, even if that were biologically possible, wouldn’t
allow you new achievements, because the many complex, sophisticated skills
we each display (including creativity, intuition, and more) depend on the
whole brain. In other words, our hemispheres are not cerebral competitors,
each trying to impose its style of thinking on the other. Instead, the hemispheres pool their specialized capacities to produce a seamlessly integrated,
single mental self.
C H A P T E R T WO The Neural Basis for Cognition
FIGURE 2.6
STUDYING SPLIT-BRAIN PATIENTS
“A fork”
When a split-brain patient is asked what he
sees, the left hemisphere sees the fork on the
right side of the screen and can verbalize that.
A
The right hemisphere sees the spoon on the
screen’s left side, but it cannot verbalize that.
However, if the patient reaches with his left hand
to pick up the object, he does select the spoon.
B
In this experiment, the patient is shown two pictures, one of a spoon and one of a fork (Panel A). If asked what
he sees, his verbal response is controlled by the left hemisphere, which has seen only the fork (because it’s in the
right visual field. However, if asked to pick up the object shown in the picture, the patient — reaching with his left
hand — picks up the spoon (Panel B). That happens because the left hand is controlled by the right hemisphere,
and this hemisphere receives visual information from the left-hand side of the visual world.
Sources of Evidence about the Brain
How can we learn about these various structures — and many others that we
haven’t named? Cognitive neuroscience relies on many types of evidence to
study the brain and nervous system. Let’s look at some of the options.
Data from Neuropsychology
We’ve already encountered one form of evidence — the study of individuals
who have suffered brain damage through accident, disease, or birth defect.
The study of these cases generally falls within the domain of neuropsychology: the study of the brain’s structures and how they relate to brain function.
Within neuropsychology, the specialty of clinical neuropsychology seeks
(among other goals) to understand the functioning of intact, undamaged
brains by means of careful scrutiny of cases involving brain damage.
Data drawn from clinical neuropsychology will be important throughout
this text. For now, though, we’ll emphasize that the symptoms resulting from
Sources of Evidence about the Brain
•
37
brain damage depend on the site of the damage. A lesion (a specific area of
damage) in the hippocampus produces memory problems but not language
disorders; a lesion in the occipital cortex produces problems in vision but
spares the other sensory modalities. Likewise, the consequences of brain
lesions depend on which hemisphere is damaged. Damage to the left side of
the frontal lobe, for example, is likely to produce a disruption of language use;
damage to the right side of the frontal lobe generally doesn’t have this effect.
In obvious ways, then, these patterns confirm the claim that different brain
areas perform different functions. In addition, these patterns provide a rich
source of data that help us develop and test hypotheses about those functions.
Data from Neuroimaging
Further insights come from neuroimaging techniques. There are several types
of neuroimaging, but they all produce precise, three-dimensional pictures of
a living brain. Some neuroimaging procedures provide structural imaging,
generating a detailed portrait of the shapes, sizes, and positions of the brain’s
components. Other procedures provide functional imaging, which tells us
about activity levels throughout the brain.
For many years, computerized axial tomography (CT scans) was the primary
tool for structural imaging, and positron emission tomography (PET scans)
was used to study the brain’s activity. CT scans rely on X-rays and so — in
essence — provide a three-dimensional X-ray picture of the brain. PET scans, in
contrast, start by introducing a tracer substance such as glucose into the patient’s
body; the molecules of this tracer have been tagged with a low dose of radioactivity, and the scan keeps track of this radioactivity, allowing us to tell which tissues
are using more of the glucose (the body’s main fuel) and which ones are using less.
For each type of scan, the primary data (X-rays or radioactive emissions)
are collected by a bank of detectors placed around the person’s head. A computer then compares the signals received by each of the detectors and uses this
information to construct a three-dimensional map of the brain — a map of
structures from a CT scan, and a map showing activity levels from a PET scan.
More recent studies have turned to two newer techniques, introduced earlier
in the chapter. Magnetic resonance imaging (MRI scans) relies on the magnetic
properties of the atoms that make up the brain tissue, and it yields fabulously
detailed pictures of the brain. MRI scans provide structural images, but a
closely related technique, functional magnetic resonance imaging (fMRI scans),
provides functional imaging. The fMRI scans measure the oxygen content in
blood flowing through each region of the brain; this turns out to be an accurate
index of the level of neural activity in that region. In this way, fMRI scans offer
an incredibly precise picture of the brain’s moment-by-moment activities.
The results of structural imaging (CT or MRI scans) are relatively stable,
changing only if the person’s brain structure changes (because of an injury,
perhaps, or the growth of a tumor). The results of PET or fMRI scans, in contrast, are highly variable, because the results depend on what task the person
is performing. We can therefore use these latter techniques to explore brain
function — using fMRI scans, for example, to determine which brain sites are
38 •
C H A P T E R T WO The Neural Basis for Cognition
PET SCANS
PET scans measure how much
glucose (the brain’s fuel) is
being used at specific locations
within the brain; this provides
a measurement of each location’s activity level at a certain
moment in time. In the figure,
the brain is viewed from above,
with the front of the head at the
top and the back of the head at
the bottom. The various colors
indicate relative activity levels
(an actual brain is uniformly
colored), using the palette
shown on the right side of the
figure. Dark blue indicates a low
level of activity; red indicates
a high level. And as the figure
shows, visual processing involves increased activity in the
occipital lobe.
especially activated when someone is making a moral judgment or trying to
solve a logic problem. In this way, the neuroimaging data can provide crucial
information about how these activities are made possible by specific patterns
of functioning within the brain.
Data from Electrical Recording
Neuroscientists have another technique in their toolkit: electrical recording of
the brain’s activity. To explain this point, though, we need to say a bit about
how the brain functions. As mentioned earlier, the brain contains billions of
nerve cells — called “neurons” — and it is the neurons that do the brain’s main
work. (We’ll say more about these cells later in the chapter.) Neurons vary in
their functioning, but for the most part they communicate with one another
via chemical signals called “neurotransmitters.” Once a neuron is “activated,”
it releases the transmitter, and this chemical can then activate (or, in some cases,
de-activate) other, adjacent neurons. The adjacent neurons “receive” this
chemical signal and, in turn, send their own signal onward to other neurons.
Let’s be clear, though, that the process we just described is communication
between neurons: One neuron releases the transmitter substance, and this activates (or de-activates) another neuron. But there’s also communication within
each neuron. The reason, basically, is that neurons have an “input” end and an
“output” end. The “input” end is the portion of the neuron that’s most sensitive
to neurotransmitters; this is where the signal from other neurons is received.
The “output” end is the portion that releases neurotransmitters, sending the
Sources of Evidence about the Brain
•
39
A
B
C
MAGNETIC RESONANCE IMAGING
Magnetic resonance imaging produces magnificently detailed pictures of the brain. Panel A shows an “axial
view” — a “slice” of the brain viewed from the top of the head (the front of the head is at the top of the image).
Clearly visible is the longitudinal fissure, which divides the left cerebral hemisphere from the right. Panel B, a “coronal view,” shows a slice of the brain viewed from the front. Again, the separation of the two hemispheres is clearly
visible, as are some of the commissures linking the two brain halves. Panel C, a “sagittal view,” shows a slice of
the brain viewed from the side. Here, many of the structures in the limbic system (see Figure 2.5) are easily seen.
signal on to other neurons. These two ends can sometimes be far apart. (For
example, some neurons in the body run from the base of the spine down to the
toes; for these cells, the input and output ends might be a full meter apart.) The
question, then, is how neurons get a signal from one end of the cell to the other.
The answer involves an electrical pulse, made possible by a flow of charged
atoms (ions) in and out of the neuron (again, we’ll say more about this process later in the chapter). The amount of electrical current involved in this ion
flow is tiny; but many millions of neurons are active at the same time, and
the current generated by all of them together is strong enough to be detected
by sensitive electrodes placed on the surface of the scalp. This is the basis
for electroencephalography — a recording of voltage changes occurring at the
scalp that reflect activity in the brain underneath. This procedure generates
an electroencephalogram (EEG) — a recording of the brain’s electrical activity.
Often, EEGs are used to study broad rhythms in the brain’s activity. For
example, an alpha rhythm (with the activity level rising and falling seven to ten
times per second) can usually be detected in the brain of someone who is awake
but calm and relaxed; a delta rhythm (with the activity rising and falling roughly
one to four times per second) is observed when someone is deeply asleep. A
much faster gamma rhythm (between 30 and 80 cycles per second) has received
a lot of research attention, with a suggestion that this rhythm plays a key role
in creating conscious awareness (e.g., Crick & Koch, 1990; Dehaene, 2014).
Sometimes, though, we want to know about the electrical activity in the
brain over a shorter period — for example, when the brain is responding to
a specific input or a particular stimulus. In this case, we measure changes in
the EEG in the brief periods just before, during, and after the event. These
changes are referred to as event-related potentials (see Figure 2.7).
40 •
C H A P T E R T WO The Neural Basis for Cognition
FIGURE 2.7
RECORDING THE BRAIN’S ELECTRICAL ACTIVITY
Alert wakefulness
Beta waves
Just before sleep
Alpha waves
Stage 1
Theta waves
Stage 2
Sleep spindle
Stage 3
K complex
Delta waves
Stage 4
A
B
EEG
–
Amplifier
+
Prestimulus
period
20 µV
Stimulus
onset
Repeat and combine
for 100 trials
ERP
–
+
Sound
generator
2 µV
Stimulus
onset
700 ms
C
To record the brain’s electrical signals, researchers generally use a cap that has electrodes attached to it. The procedure is easy and entirely safe — it can even be used to measure brain signals in a baby (Panel A). In some procedures, researchers measure recurrent rhythms in the brain’s activity, including rhythms that distinguish the stages
of sleep (Panel B). In other procedures, they measure brain activity produced in response to a single event — such
as the presentation of a well-defined stimulus (Panel C).
The Power of Combining Techniques
Each of the research tools we’ve described has strengths and weaknesses. CT
scans and MRI data tell us about the shape and size of brain structures, but
they tell nothing about the activity levels within these structures. PET scans
and fMRI studies do tell us about brain activity, and they can locate the activity rather precisely (within a millimeter or two). But these techniques are less
precise about when the activity took place. For example, fMRI data summarize the brain’s activity over a period of several seconds and cannot indicate
when exactly, within this time window, the activity took place. EEG data give
more precise information about timing but are much weaker in indicating
where the activity took place.
Researchers deal with these limitations by means of a strategy commonly
used in science: We seek data from multiple sources, so that the strengths of
one technique can make up for the shortcomings of another. As a result, some
studies combine EEG recordings with fMRI scans, with the EEGs telling us
when certain events took place in the brain, and the scans telling us where
the activity took place. Likewise, some studies combine fMRI scans with
CT data, so that findings about brain activation can be linked to a detailed
portrait of the person’s brain anatomy.
Researchers also face another complication: the fact that many of the
techniques described so far provide correlational data. To understand
the concern here, let’s look at an example. A brain area called the fusiform face area (FFA) is especially active whenever a face is being perceived
(see Figure 2.8) — and so there is a correlation between a mental activity
(perceiving a face) and a pattern of brain activity. Does this mean the FFA
is needed for face perception? A different possibility is that the FFA activation may just be a by-product of face perception and doesn’t play a crucial
role. As an analogy, think about the fact that a car’s speedometer becomes
“more activated” (i.e., shows a higher value) whenever the car goes faster.
That doesn’t mean that the speedometer causes the speed or is necessary
for the speed. The car would go just as fast and would, for many purposes,
perform just as well if the speedometer were removed. The speedometer’s
state, in other words, is correlated with the car’s speed but in no sense
causes (or promotes, or is needed for) the car’s speed.
In the same way, neuroimaging data can tell us that a brain area’s activity
is correlated with a particular function, but we need other data to determine
whether the brain site plays a role in causing (or supporting, or allowing)
that function. In many cases, those other data come from the study of brain
lesions. If damage to a brain site disrupts a particular function, it’s an indication that the site does play some role in supporting that function. (And, in
fact, the FFA does play an important role in face recognition.)
Also helpful here is a technique called transcranial magnetic stimulation
(TMS). This technique creates a series of strong magnetic pulses at a specific
location on the scalp, and these pulses activate the neurons directly underneath this scalp area (Helmuth, 2001). TMS can thus be used as a means of
42 •
C H A P T E R T WO The Neural Basis for Cognition
FIGURE 2.8
BRAIN ACTIVITY AND AWARENESS
FFA
A
Percentage of fMRI signal
Face
Face
1.0
FFA
0.8
House
0.8
PPA
0.6
0.6
PPA
0.4
0.4
FFA
0.2
0.2
0.0
C
B
House
1.0
PPA
PPA
–8
–4
0
4
8
12
0.0
–8
–4
0
4
8
12
Time from reported perceptual switch (s)
Panel A shows an fMRI scan of a subject looking at faces. Activation levels are high in the fusiform face area (FFA),
an area that is apparently more responsive to faces than to other visual stimuli. Panel B shows a scan of the same
subject looking at pictures of places; now, activity levels are high in the parahippocampal place area (PPA). Panel
C compares the activity in these two areas when the subject has a picture of a face in front of one eye and a picture of a house in front of the other eye. When the viewer’s perception shifts from the house to the face, activation
increases in the FFA. When the viewer’s perception shifts from the face to the house, PPA activation increases. In
this way, the activation level reflects what the subject is aware of, and not just the pattern of incoming stimulation.
( after tong , nakayama , vaughan , & kanwisher , 1998)
asking what happens if we stimulate certain neurons. In addition, because
this stimulation disrupts the ordinary function of these neurons, it produces
a (temporary) lesion — allowing us to identify, in essence, what functions are
compromised when a particular bit of brain tissue is briefly “turned off.” In
these ways, the results of a TMS procedure can provide crucial information
about the functional role of that brain area.
Sources of Evidence about the Brain
•
43
Localization of Function
TEST YOURSELF
5.What is the difference
between structural imaging of the brain and
functional imaging?
What techniques are
used for each?
6.What do we gain from
combining different
methods in studying
the brain?
7.What is meant by the
phrase “localization
of function”?
Drawing on the techniques we have described, neuroscientists have learned
a great deal about the function of specific brain structures. This type of
research effort is referred to as the localization of function, an effort (to put
it crudely) aimed at figuring out what’s happening where within the brain.
Localization data are useful in many ways. For example, think back to
the discussion of Capgras syndrome earlier in this chapter. Brain scans told
us that people with this syndrome have damaged amygdalae, but how is this
damage related to the symptoms of the syndrome? More broadly, what problems does a damaged amygdala create? To tackle these questions, we rely on
localization of function — in particular, on data showing that the amygdala
is involved in many tasks involving emotional appraisal. This combination
of points helped us to build (and test) our claims about this syndrome and,
in general, claims about the role of emotion within the ordinary experience
of “familiarity.”
As a different illustration, consider the experience of calling up a “mental
picture” before the “mind’s eye.” We’ll have more to say about this experience
in Chapter 11, but we can already ask: How much does this experience have
in common with ordinary seeing — that is, the processes that unfold when
we place a real picture before someone’s eyes? As it turns out, localization
data reveal enormous overlap between the brain structures needed for these
two activities (visualizing and actual vision), telling us immediately that these
activities do have a great deal in common (see Figure 2.9). So, again, we build
on localization — this time to identify how exactly two mental activities are
related to each other.
The Cerebral Cortex
As we’ve noted, the largest portion of the human brain is the cerebral cortex —
the thin layer of tissue covering the cerebrum. This is the region in which an
enormous amount of information processing takes place, and so, for many
topics, it is the brain region of greatest interest for cognitive psychologists.
The cortex includes many distinct regions, each with its own function,
but these regions are traditionally divided into three categories. Motor areas
contain brain tissue crucial for organizing and controlling bodily movements. Sensory areas contain tissue essential for organizing and analyzing
the information received from the senses. Association areas support many
functions, including the essential (but not well-defined) human activity we
call “thinking.”
Motor Areas
Certain regions of the cerebral cortex serve as the “departure points” for
signals leaving the cortex and controlling muscle movement. Other areas
are the “arrival points” for information coming from the eyes, ears, and
44 •
C H A P T E R T WO The Neural Basis for Cognition
FIGURE 2.9
A PORTRAIT OF THE BRAIN AT WORK
Activity while
looking at
pictures
Activity while
visualizing “mental
pictures”
These fMRI images show different “slices” through the living brain, revealing levels of activity in different brain sites. Regions that are more active are
shown in yellow, orange, and red; lower activity levels are indicated in blue.
The first column shows brain activity while a person is making judgments
about simple pictures. The second column shows brain activity while the person is making the same sorts of judgments about “mental pictures,” visualized
before the “mind’s eye.”
The Cerebral Cortex
•
45
other sense organs. In both cases, these areas are called “primary projection
areas,” with the departure points known as the primary motor projection
areas and the arrival points contained in regions known as the primary sensory
projection areas.
Evidence for the motor projection area comes from studies in which
investigators apply mild electrical current to this area in anesthetized animals.
This stimulation often produces specific movements, so that current applied
to one site causes a movement of the left front leg, while current applied to
a different site causes the ears to prick up. These movements show a pattern
of contralateral control, with stimulation to the left hemisphere leading to
movements on the right side of the body, and vice versa.
Why are these areas called “projection areas”? The term is borrowed from
mathematics and from the discipline of map making, because these areas
seem to form “maps” of the external world, with particular positions on the
cortex corresponding to particular parts of the body or particular locations
in space. In the human brain, the map that constitutes the motor projection
area is located on a strip of tissue toward the rear of the frontal lobe, and the
pattern of mapping is illustrated in Figure 2.10. In this illustration, a drawing
of a person has been overlaid on a depiction of the brain, with each part of
the little person positioned on top of the brain area that controls its movement. The figure shows that areas of the body that we can move with great
precision (e.g., fingers and lips) have a lot of cortical area devoted to them;
areas of the body over which we have less control (e.g., the shoulder and the
back) receive less cortical coverage.
Sensory Areas
Information arriving from the skin senses (your sense of touch or your sense
of temperature) is projected to a region in the parietal lobe, just behind the
motor projection area. This is labeled the “somatosensory” area in Figure 2.10.
If a patient’s brain is stimulated in this region (with electrical current or touch),
the patient will typically report a tingling sensation in a specific part of the
body. Figure 2.10 also shows the region (in the temporal lobes) that functions
as the primary projection area for hearing (the “auditory” area). If the brain is
directly stimulated here, the patient will hear clicks, buzzes, and hums. An area
in the occipital lobes is the primary projection area for vision; stimulation here
causes the patient to see flashes of light or visual patterns.
The sensory projection areas differ from each other in important ways,
but they also have features in common — and they’re features that parallel
the attributes of the motor projection area. First, each of these areas provides
a “map” of the sensory environment. In the somatosensory area, each part
of the body’s surface is represented by its own region on the cortex; areas of
the body that are near to each other are typically represented by similarly
nearby areas in the brain. In the visual area, each region of visual space has
its own cortical representation, and adjacent areas of visual space are usually
represented by adjacent brain sites. In the auditory projection area, different
46 •
C H A P T E R T WO The Neural Basis for Cognition
Tongue
Primary auditory
cortex
Th
u
Ey mb
e
N
Fa os
ce e
Hand
Fingers
Genitals
Lips
Gums
Teeth
Jaw
Hip
Trunk
Neck
Head
Shoulder
Arm
Arm
Shoulder
Neck
Trunk
Hip
Fingers
Hand
bk
umec
Th N rowe
B Ey
e
os ce
Fa
N
Knee Knee
Leg
Leg
Ankle Foot
Toes Toes
FIGURE 2.10 THE PRIMARY
PROJECTION AREAS
Primary motor
projection area
Lips
Gums
Teeth
Jaw
Primary
somatosensory
projection area
Tongue
Primary visual
cortex
The primary motor projection area is
located at the rearmost edge of the
frontal lobe, and each region within
this projection area controls the
motion of a specific body part, as
illustrated on the top left. The primary
somatosensory projection area, receiving information from the skin, is
at the forward edge of the parietal
lobe; each region within this area
receives input from a specific body
part. The primary projection areas
for vision and hearing are located
in the occipital and temporal lobes,
respectively. These two areas are
also organized systematically. For
example, in the visual projection
area, adjacent areas of the brain
receive visual inputs that come from
adjacent areas in visual space.
frequencies of sound have their own cortical sites, and adjacent brain sites
are responsive to adjacent frequencies.
Second, in each of these sensory maps, the assignment of cortical space is
governed by function, not by anatomical proportions. In the parietal lobes,
parts of the body that aren’t very discriminating with regard to touch — even
if they’re physically large — get relatively little cortical area. Other, more
sensitive areas of the body (the lips, tongue, and fingers) get much more
space. In the occipital lobes, more cortical surface is devoted to the fovea,
the part of the eyeball that is most sensitive to detail. (For more on the
fovea, see Chapter 3.) And in the auditory areas, some frequencies of sound
get more cerebral coverage than others. It’s surely no coincidence that these
“advantaged” frequencies are those essential for the perception of speech.
Finally, we also find evidence here of contralateral connections. The
somatosensory area in the left hemisphere, for example, receives its main
input from the right side of the body; the corresponding area in the right
hemisphere receives its input from the left side of the body. Likewise for
the visual projection areas, although here the projection is not contralateral
with regard to body parts. Instead, it’s contralateral with regard to physical
space. Specifically, the visual projection area in the right hemisphere receives
information from both the left eye and the right, but the information it
receives corresponds to the left half of visual space (i.e., all of the things
The Cerebral Cortex
•
47
visible to your left when you’re looking straight ahead). The reverse is true
for the visual area in the left hemisphere. It receives information from both
eyes, but from only the right half of visual space. The pattern of contralateral
organization is also evident — although not as clear-cut — for the auditory
cortex, with roughly 60% of the nerve fibers from each ear sending their
information to the opposite side of the brain.
Association Areas
THE SENSORY
HOMUNCULUS
An artist’s rendition of what
a man would look like if his
appearance were proportional
to the area allotted by the
somatosensory cortex to his
various body parts.
TEST YOURSELF
8.What is a projection
area in the brain?
What’s the role of the
motor projection area?
The sensory projection
area?
9.What does it mean
to say that the brain
relies on “contralateral”
connections?
48 •
The areas described so far, both motor and sensory, make up only a small
part of the human cerebral cortex — roughly 25%. The remaining cortical areas are traditionally referred to as the association cortex. This terminology is falling out of use, however, partly because this large volume of
brain tissue can be subdivided further on both functional and anatomical
grounds. These subdivisions are perhaps best revealed by the diversity of
symptoms that result if the cortex is damaged in one or another specific
location. For example, some lesions in the frontal lobe produce apraxias,
disturbances in the initiation or organization of voluntary action. Other
lesions (generally in the occipital cortex, or in the rearmost part of the
parietal lobe) lead to agnosias, disruptions in the ability to identify familiar
objects. Agnosias usually affect one modality only — so a patient with visual
agnosia, for example, can recognize a fork by touching it but not by looking
at it. A patient with auditory agnosia, by contrast, might be unable to identify familiar voices but might still recognize the face of the person speaking.
Still other lesions (usually in the parietal lobe) produce neglect syndrome, in
which the individual seems to ignore half of the visual world. A patient afflicted
with this syndrome will shave only half of his face and eat food from only half
of his plate. If asked to read the word “parties,” he will read “ties,” and so on.
Damage in other areas causes still other symptoms. We mentioned earlier
that lesions in areas near the lateral fissure (the deep groove that separates the
frontal and temporal lobes) can result in disruption to language capacities, a
problem referred to as aphasia.
Finally, damage to the frontmost part of the frontal lobe, the prefrontal area, causes problems in planning and implementing strategies. In some
cases, patients with damage here show problems in inhibiting their own
behaviors, relying on habit even in situations for which habit is inappropriate.
Frontal lobe damage can also (as we mentioned in our discussion of Capgras
syndrome) lead to a variety of confusions, such as whether a remembered
episode actually happened or was simply imagined.
We’ll discuss more about these diagnostic categories — aphasia, agnosia,
neglect, and more — in upcoming chapters, where we’ll consider these disorders in the context of other things that are known about object recognition,
attention, and so on. Our point for the moment, though, is simple: These
clinical patterns make it clear that the so-called association cortex contains
many subregions, each specialized for a particular function, but with all of
the subregions working together in virtually all aspects of our daily lives.
C H A P T E R T WO The Neural Basis for Cognition
Brain Cells
Our brief tour so far has described some of the large-scale structures in the
brain. For many purposes, though, we need to zoom in for a closer look, in
order to see how the brain’s functions are actually carried out.
Neurons and Glia
We’ve already mentioned that the human brain contains many billions of
neurons and a comparable number of glia. The glia perform many functions.
They help to guide the development of the nervous system in the fetus and
young infant; they support repairs if the nervous system is damaged; they
also control the flow of nutrients to the neurons. Specialized glial cells also
provide a layer of electrical insulation surrounding parts of some neurons;
this insulation dramatically increases the speed with which neurons can send
their signals. (We’ll return to this point in a moment.) Finally, some research
suggests the glia may also constitute their own signaling system within
the brain, separate from the information flow provided by the neurons
(e.g., Bullock et al, 2005; Gallo & Chitajullu, 2001).
There is no question, though, that the main flow of information through
the brain — from the sense organs inward, from one part of the brain to
the others, and then from the brain outward — is made possible by the
neurons. Neurons come in many shapes and sizes (see Figure 2.11), but
in general, neurons have three major parts. The cell body is the portion of
the cell that contains the neuron’s nucleus and all the elements needed for
the normal metabolic activities of the cell. The dendrites are usually the
FIGURE 2.11
A
NEURONS
B
C
Panel A shows neurons from the spinal cord (stained in red); Panel B shows neurons from the cerebellum; Panel C
shows neurons from the cerebral cortex.
Brain Cells
•
49
“input” side of the neuron, receiving signals from many other neurons. In
most neurons, the dendrites are heavily branched, like a thick and tangled
bush. The axon is the “output” side of the neuron; it sends neural impulses to other neurons (see Figure 2.12). Axons can vary enormously in
length — the giraffe, for example, has neurons with axons that run the full
length of its neck.
FIGURE 2.12
REGIONS OF THE NEURON
Dendrites
Nucleus
Cell body
Myelin sheath
Axon
Neural
impulse
Nodes of Ranvier
Axon terminals
Most neurons have three identifiable regions. The dendrites are the part of
the neuron that usually detects incoming signals. The cell body contains the
metabolic machinery that sustains the cell. The axon is the part of the neuron
that transmits a signal to another location. When the cell fires, neurotransmitters are released from the terminal endings at the tip of the axon. The myelin
sheath is created by glial cells that wrap around the axons of many neurons.
The gaps in between the myelin cells are called the nodes of Ranvier.
50 •
C H A P T E R T WO The Neural Basis for Cognition
COGNITION
outside the lab
Alcohol
Of the many drugs that influence the brain, one
factors also matter. For example, blackouts are
is readily available and often consumed: alcohol.
more common if you become drunk rapidly — as
Alcohol influences the entire brain, and even at
when you drink on an empty stomach, or when
low levels of intoxication we can detect alcohol’s
you gulp alcohol rather than sipping it.
effects — for example, with measures of motor
skills or response time.
Let’s combine these points about blackouts,
though, with our earlier observation about alco-
Alcohol’s effects are more visible, though, in
hol’s uneven effects. It’s possible for someone to
some functions than in others, and so someone
be quite drunk, and therefore suffer an alcoholic
who’s quite intoxicated can perform many activi-
blackout, even if the person seemed alert and
ties at a fairly normal level. However, alcohol has
coherent during the drunken episode. To see
a strong impact on activities that depend on the
some of the serious problems this can cause,
brain’s prefrontal cortex. This is the brain region
consider a pattern that often emerges in cases
that’s essential for the mind’s executive function —
involving allegations of sexual assault. Victims of
the system that allows you to control your thoughts
assault sometimes report that they have little or
and behaviors. (We’ll say more about executive
no memory of the sexual encounter; they there-
function in upcoming chapters.) As a result, alco-
fore assume they were barely conscious during
hol undercuts your ability to resist temptation or
the event and surely incapable of giving consent.
to overcome habit. Impairments in executive func-
But is this assumption correct?
tion also erode your ability to make thoughtful
decisions and draw sensible conclusions.
The answer is complex. If someone was drunk
enough to end up with a blackout, then that per-
In addition, alcohol can produce impair-
son was probably impaired to a degree that would
ments in memory, including “alcoholic blackouts.”
interfere with decision making — and so the per-
So-called fragmentary blackouts, in which the
son could not have given legitimate, meaningful
person remembers some bits of an experience
consent. But, even so, the person might have been
but not others, are actually quite common. In one
functioning in a way that seemed mostly normal
study, college students were asked: “Have you
(able to converse, to move around) and may even
ever awoken after a night of drinking not able to
have expressed consent in words or actions.
remember things that you did or places where you
In this situation, then, the complainant is cor-
went?” More than half of the students indicated that,
rect in saying that he or she couldn’t have given
yes, this had happened to them at some point;
(and therefore didn’t give) meaningful consent,
40% reported they’d had a blackout within the
but the accused person can legitimately say that
previous year.
he or she perceived that there was consent. We
How drunk do you have to be in order to
can debate how best to judge these situations, but
experience a blackout? Many authorities point to
surely the best path forward is to avoid this sort of
a blood alcohol level of 0.25 (roughly nine or ten
circumstance — by drinking only in safe settings or
drinks for someone of average weight), but other
by keeping a strict limit on your drinking.
Brain Cells
•
51
The Synapse
We’ve mentioned that communication from one neuron to the next is generally made possible by a chemical signal: When a neuron has been sufficiently
stimulated, it releases a minute quantity of a neurotransmitter. The molecules
of this substance drift across the tiny gap between neurons and latch on
to the dendrites of the adjacent cell. If the dendrites receive enough of this
substance, the next neuron will “fire,” and so the signal will be sent along to
other neurons.
Notice, then, that neurons usually don’t touch each other directly. Instead,
at the end of the axon there is a gap separating each neuron from the next.
This entire site — the end of the axon, plus the gap, plus the receiving membrane of the next neuron — is called a synapse. The space between the neurons
is the synaptic gap. The bit of the neuron that releases the transmitter into this
gap is the presynaptic membrane, and the bit of the neuron on the other side
of the gap, affected by the transmitters, is the postsynaptic membrane.
When the neurotransmitters arrive at the postsynaptic membrane, they
cause changes in this membrane that enable certain ions to flow into and out
of the postsynaptic cell (see Figure 2.13). If these ionic flows are relatively
small, then the postsynaptic cell quickly recovers and the ions are transported
back to where they were initially. But if the ionic flows are large enough, they
trigger a response in the postsynaptic cell. In formal terms, if the incoming
signal reaches the postsynaptic cell’s threshold, then the cell fires. That is, it
produces an action potential — a signal that moves down its axon, which in
turn causes the release of neurotransmitters at the next synapse, potentially
causing the next cell to fire.
In some neurons, the action potential moves down the axon at a relatively
slow speed. For other neurons, specialized glial cells are wrapped around the
axon, creating a layer of insulation called the myelin sheath (see Figure 2.12).
Because of the myelin, ions can flow in or out of the axon only at the gaps
between the myelin cells. As a result, the signal traveling down the axon has
to “jump” from gap to gap, and this greatly increases the speed at which the
signal is transmitted. For neurons without myelin, the signal travels at speeds
below 10 m/s; for “myelinated” neurons, the speed can be ten times faster.
Overall, let’s emphasize four points about this sequence of events. First,
let’s note once again that neurons depend on two different forms of information flow. Communication from one neuron to the next is (for most neurons)
mediated by a chemical signal. In contrast, communication from one end of
the neuron to the other (usually from the dendrites down the length of the
axon) is made possible by an electrical signal, created by the flow of ions in
and out of the cell.
Second, the postsynaptic neuron’s initial response can vary in size; the
incoming signal can cause a small ionic flow or a large one. Crucially, though,
once these inputs reach the postsynaptic neuron’s firing threshold, there’s no
variability in the response — either a signal is sent down the axon or it is not.
If the signal is sent, it is always of the same magnitude, a fact referred to as
52 •
C H A P T E R T WO The Neural Basis for Cognition
FIGURE 2.13
SCHEMATIC VIEW OF SYNAPTIC TRANSMISSION
Neuron 1
A
Action
potential
Neuron 2
Axon
Synaptic
vesicle
Synaptic
vesicle
Synaptic
gap
Na+
Ion channel
C
Receptor site
Neurotransmitter
Postsynaptic
membrane
Dendrite
B
Presynaptic
membrane
Na+
D
(Panel A) Neuron 1 transmits a message across the synaptic gap to Neuron 2. The neurotransmitters are initially
stored in structures called “synaptic vesicles” (Panel B). When a signal travels down the axon, the vesicles are
stimulated and some of them burst (Panel C), ejecting neurotransmitter molecules into the synaptic gap and
toward the postsynaptic membrane (Panel D). Neurotransmitter molecules settle on receptor sites, ion channels
open, and sodium (Na+) floods in.
Brain Cells
•
53
TEST YOURSELF
10.What are glia? What
are dendrites? What
is an axon? What is a
synapse?
11.What does it mean
to say that neurons
rely on two different
forms of information
flow, one chemical
and one electrical?
the all-or-none law. Just as pounding on a car horn won’t make the horn
any louder, a stronger stimulus won’t produce a stronger action potential. A
neuron either fires or it doesn’t; there’s no in-between.
This does not mean, however, that neurons always send exactly the same
information. A neuron can fire many times per second or only occasionally. A
neuron can fire just once and then stop, or it can keep firing for an extended span.
But, even so, each individual response by the neuron is always the same size.
Third, we should also note that the brain relies on many different neurotransmitters. By some counts, a hundred transmitters have been catalogued
so far, and this diversity enables the brain to send a variety of different messages. Some transmitters have the effect of stimulating subsequent neurons;
some do the opposite and inhibit other neurons. Some transmitters play an
essential role in learning and memory; others play a key role in regulating the
level of arousal in the brain; still others influence motivation and emotion.
Fourth, let’s be clear about the central role of the synapse. The synaptic gap is actually quite small — roughly 20 to 30 nanometers across. (For
contrast’s sake, the diameter of a human hair is roughly 80,000 nano­
meters.) Even so, transmission across this gap slows down the neuronal
signal, but this is a tiny price to pay for the advantages created by this mode
of signaling: Each neuron receives information from (i.e., has synapses with)
many other neurons, and this allows the “receiving” neuron to integrate
information from many sources. This pattern of many neurons feeding into
one also makes it possible for a neuron to “compare” signals and to adjust its
response to one input according to the signal arriving from a different input.
In addition, communication at the synapse is adjustable. This means that
the strength of a synaptic connection can be altered by experience, and this
adjustment is crucial for the process of learning — the storage of new knowledge and new skills within the nervous system.
Coding
This discussion of individual neurons leads to a further question: How do
these microscopic nerve cells manage to represent a specific idea or a specific
content? Let’s say that right now you’re thinking about your favorite song.
How is this information represented by neurons? The issue here is referred
to as coding, and there are many options for what the neurons’ “code” might
be (Gallistel, 2017). As one option, we might imagine that a specific group of
neurons somehow represents “favorite song,” so that whenever you’re thinking about the song, it’s precisely these neurons that are activated. Or, as a different option, the song might be represented by a broad pattern of neuronal
activity. If so, “favorite song” might be represented in the brain by something
like “Neuron X firing strongly while Neuron Y is firing weakly and Neuron
Z is not firing at all” (and so on for thousands of other neurons). Note that
within this scheme the same neurons might be involved in the representation
of other sounds, but with different patterns. So — to continue our example —
Neuron X might also be involved in the representation of the sound of a car
54 •
C H A P T E R T WO The Neural Basis for Cognition
engine, but for this sound it might be part of a pattern that includes Neurons
Q, R, and S also firing strongly, and Neuron Y not firing at all.
As it turns out, the brain uses both forms of coding. For example, in
Chapter 4 we’ll see that some neurons really are associated with a particular
content. In fact, researchers documented a cell in one of the people they tested
that fired whenever a picture of Jennifer Aniston was in view, and didn’t fire in
response to pictures of other faces. Another cell (in a different person’s brain)
fired whenever a picture of the Sydney Opera House was shown, but didn’t
fire when other buildings were in view (Quiroga, Reddy, Kreiman, Koch, &
Fried, 2005)! These do seem to be cases in which an idea (in particular, a
certain visual image) is represented by specific neurons in the brain.
In other cases, evidence suggests that ideas and memories are represented
in the brain through widespread patterns of activity. This sort of “pattern
coding” is, for example, certainly involved in the neural mechanisms through
which you plan, and then carry out, particular motions — like reaching out to
turn a book page or lifting your foot to step over an obstacle (Georgopoulos,
1990, 1995). We’ll return to pattern coding in Chapter 9, when we discuss
the notion of a distributed representation.
Moving On
We have now described the brain’s basic anatomy and have also taken a brief
look at the brain’s microscopic parts — the individual neurons. But how do
all of these elements, large and small, function in ways that enable us to think,
remember, learn, speak, or feel? As a step toward tackling this issue, the next
chapter takes a closer look at the portions of the nervous system that allow us
to see. We’ll use the visual system as our example for two important reasons.
First, vision is the modality through which humans acquire a huge amount of
information, whether by reading or simply by viewing the world around us.
If we understand vision, therefore, we understand the processes that bring us
much of our knowledge. Second, investigators have made enormous progress
in mapping out the neural “wiring” of the visual system, offering a detailed
and sophisticated portrait of how this system operates. As a result, an
examination of vision provides an excellent illustration of how the study of
the brain can proceed and what it can teach us.
TEST YOURSELF
12. H
ow is information
coded, or represented,
in the brain?
COGNITIVE PSYCHOLOGY AND EDUCATION
food supplements and cognition
Various businesses try to sell you training programs or food supplements that
(they claim) will improve your memory, help you think more clearly, and so
on. Evidence suggests, though, that the currently offered training programs
Cognitive Psychology and Education
•
55
GINKGO BILOBA
A variety of food supplements
derived from the ginkgo tree
are alleged to improve cognitive functioning. Current understanding, though, suggests
that the benefits of Ginkgo
biloba are indirect: This supplement improves functioning
because it can improve blood
circulation and can help the
body to fight some forms of
inflammation.
56 •
may provide little benefit. These programs do improve performance on the
specific exercises contained within the training itself, but they have no impact
on any tasks beyond these exercises. In other words, the programs don’t seem
to help with the sorts of mental challenges you encounter in day-to-day functioning (Simons et al., 2016).
What about food supplements? Most of these supplements have not been
tested in any systematic way, and so there’s little (and often no) solid evidence
to support the claims sometimes made for these products. One supplement,
though, has been rigorously tested: Ginkgo biloba, an extract derived from
a tree of the same name and advertised as capable of enhancing memory. Is
Ginkgo biloba effective? To answer that question, let’s begin with the fact
that for its normal functioning, the brain requires an excellent blood flow
and, with that, a lot of oxygen and a lot of nutrients. Indeed, it’s estimated
that the brain, constituting roughly 2% of your body weight, consumes
15% percent of your body’s energy supply.
It’s not surprising, therefore, that the brain’s operations are impaired if
some change in your health interferes with the flow of oxygen or nutrients.
If (for example) you’re ill, or not eating enough, or not getting enough sleep,
these conditions affect virtually all aspects of your biological functioning.
However, since the brain is so demanding of nutrients and oxygen, it’s one
of the first organs to suffer if the supply of these necessities is compromised.
This is why poor nutrition or poor health almost inevitably undermines your
ability to think, to remember, or to pay attention.
Within this context, it’s important that Ginkgo biloba can improve blood
circulation and reduce some sorts of bodily inflammation. Because of these
effects, Ginkgo can be helpful for people who have circulatory problems or
who are at risk for nerve damage, and one group that may benefit is patients
with Alzheimer’s disease. Evidence suggests that Ginkgo helps these patients
remember more and think more clearly, but this isn’t because Ginkgo is making these patients “smarter” in any direct way. Instead, the Ginkgo is broadly
improving the patients’ blood circulation and the health status of their nerve
cells, allowing these cells to do their work.
What about healthy people — those not suffering from bodily inflammations or damage to their brain cells? Here, the evidence is mixed, but most
studies have observed no benefit from this food supplement. Apparently,
Ginkgo’s effects, if they exist at all in healthy adults, are so small that they’re
difficult to detect.
Are there other steps that will improve the mental functioning of healthy
young adults? Answers here have to be tentative, because new “smart pills”
and “smart foods” are being proposed all the time, and each one has to be
tested before we can know its effects. For now, though, we’ve already indicated part of a positive answer: Good nutrition, plenty of sleep, and adequate
exercise will keep your blood supply in good condition, and this will help
your brain to do its job. In addition, there may be something else you can do.
The brain needs “fuel” to do its work, and the body’s fuel comes from the
sugar glucose. You can protect yourself, therefore, by making sure that your
C H A P T E R T WO The Neural Basis for Cognition
brain has all the glucose it needs. This isn’t a recommendation to jettison all
other aspects of your diet and eat nothing but chocolate bars. In fact, most
of the glucose your body needs doesn’t come from sugary foods; instead,
most comes from the breakdown of carbohydrates — from the grains, dairy
products, fruits, and vegetables you eat. For this reason, it might be a good
idea to have a slice of bread and a glass of milk just before taking an exam or
walking into a particularly challenging class. These steps will help make sure
that you’re not caught by a glucose shortfall that could interfere with your
brain’s functioning.
Also, be careful not to ingest too much sugar. If you eat a big candy bar
just before an exam, you might get an upward spike in your blood glucose
followed by a sudden drop, and these abrupt changes can produce problems
of their own.
Overall, then, it seems that food supplements tested so far offer no “fast
track” toward better cognition. Ginkgo biloba is helpful, but mostly for special populations. A high-carb snack may help, but it will be of little value if
you’re already adequately nourished. Therefore, on all these grounds, the
best path toward better cognition seems to be the one that common sense
would already recommend — eating a balanced diet, getting a good night’s
sleep, and paying careful attention during your studies.
For more on this topic . . .
Allen, A. L., & Strand, N. K. (2015). Cognitive enhancement and beyond:
Recommendations from the Bioethics Commission. Trends in Cognitive
Sciences, 19, 549–555.
Gold, P. E., Cahill, L., & Wenk, G. L. (2002). Ginkgo biloba: A cognitive
enhancer? Psychological Science in the Public Interest, 3, 2–11.
Husain, M., & Mehta, M. A. (2011). Cognitive enhancement by drugs in health
and disease. Trends in Cognitive Sciences, 15, 28–36.
Masicampo, E., & Baumeister, R. (2008). Toward a physiology of dual-process
reasoning and judgment: Lemonade, willpower, and expensive rule-based
analysis. Psychological Science, 19, 255–260.
McDaniel, M. A., Maier, S. F., & Einstein, G. O. (2002). “Brain-specific” nutrients:
A memory cure? Psychological Science in the Public Interest, 3, 12–38.
Stough, C., & Pase, M. P. (2015). Improving cognition in the elderly with
nutritional supplements. Current Directions in Psychological Sciences, 24,
177–183.
Cognitive Psychology and Education
•
57
chapter review
SUMMARY
• The brain is divided into several different structures, but of particular importance for cognitive
psychology is the forebrain. In the forebrain, each
cerebral hemisphere is divided into the frontal lobe,
parietal lobe, temporal lobe, and occipital lobe. In
understanding these brain areas, one important
source of evidence comes from studies of brain damage, enabling us to examine what sorts of symptoms
result from lesions in specific brain locations. This
has allowed a localization of function, an effort that
is also supported by neuroimaging research, which
shows that the pattern of activation in the brain
depends on the particular task being performed.
• Different parts of the brain perform different jobs;
but for virtually any mental process, different brain
areas must work together in a closely integrated way.
When this integration is lost (as it is, for example,
in Capgras syndrome), bizarre symptoms can result.
• The primary motor projection areas are the
departure points in the brain for nerve cells that
initiate muscle movement. The primary sensory projection areas are the main points of arrival in the
brain for information from the eyes, ears, and other
sense organs. These projection areas generally show
a pattern of contralateral control, with tissue in the
left hemisphere sending or receiving its main signals
from the right side of the body, and vice versa. Each
projection area provides a map of the environment
or the relevant body part, but the assignment of
space in this map is governed by function, not by
anatomical proportions.
• Most of the forebrain’s cortex has traditionally
been referred to as the association cortex, but this
area is subdivided into specialized regions. This
subdivision is reflected in the varying consequences
of brain damage, with lesions in the occipital lobes
leading to visual agnosia, damage in the temporal
lobes leading to aphasia, and so on. Damage to the
prefrontal area causes many different problems, but
these are generally in the forming and implementing
of strategies.
• The brain’s functioning depends on neurons and
glia. The glia perform many functions, but the main
flow of information is carried by the neurons. Communication from one end of the neuron to the other
is electrical and is governed by the flow of ions in
and out of the cell. Communication from one neuron to the next is generally chemical, with a neuron
releasing neurotransmitters that affect neurons on
the other side of the synapse.
KEY TERMS
amygdala (p. 28)
prefrontal cortex (p. 29)
hindbrain (p. 32)
cerebellum (p. 33)
midbrain (p. 33)
forebrain (p. 34)
cortex (p. 34)
convolutions (p. 34)
58
longitudinal fissure (p. 34)
cerebral hemisphere (p. 34)
frontal lobes (p. 34)
central fissure (p. 34)
parietal lobes p. 34)
lateral fissure (p. 34)
temporal lobes (p. 34)
occipital lobes (p. 34)
subcortical structures (p. 34)
thalamus (p. 34)
hypothalamus (p. 34)
limbic system (p. 34)
hippocampus (p. 34)
commisures (p. 36)
corpus callosum (p. 36)
lesion (p. 38)
neuroimaging techniques (p. 38)
computerized axial tomography
(CT scans) (p. 38)
positron emission tomography
(PET scans) (p. 38)
magnetic resonance imaging
(MRI scans) (p. 38)
functional magnetic resonance imaging
(fMRI scans) (p. 38)
electroencephalogram (EEG) (p. 40)
event-related potentials (p. 40)
fusiform face area (FFA) (p. 42)
transcranial magnetic stimulation (TMS) (p. 42)
localization of function (p. 44)
primary motor projection areas (p. 46)
primary sensory projection areas (p. 46)
contralateral control (p. 46)
association cortex (p. 48)
apraxias (p. 48)
agnosias (p. 48)
neglect syndrome (p. 48)
aphasia (p. 48)
neurons (p. 49)
glia (p. 49)
cell body (p. 49)
dendrites (p. 49)
axon (p. 50)
neurotransmitter (p. 52)
synapse (p. 52)
presynaptic membrane (p. 52)
postsynaptic membrane (p. 52)
threshold (p. 52)
action potential (p. 52)
myelin sheath (p. 52)
all-or-none law (p. 54)
coding (p. 54)
TEST YOURSELF AGAIN
1.What are the symptoms of Capgras syndrome,
and why do they suggest a two-part explanation
for how people recognize faces?
8.What is a projection area in the brain? What’s
the role of the motor projection area? The
sensory projection area?
2. What is the cerebral cortex?
3. What are the four major lobes of the forebrain?
9.What does it mean to say that the brain relies
on “contralateral” connections?
4.Identify some of the functions of the hippocampus, the amygdala, and the corpus callosum.
10.What are glia? What are dendrites? What is an
axon? What is a synapse?
5.What is the difference between structural
imaging of the brain and functional imaging?
What techniques are used for each?
11.What does it mean to say that neurons rely on
two different forms of information flow, one
chemical and one electrical?
6.What do we gain from combining different
methods in studying the brain?
12.How is information coded, or represented, in
the brain?
7.What is meant by the phrase “localization of
function”?
59
THINK ABOUT IT
1.People often claim that humans only use 10%
of their brains. Does anything in this chapter
help us in evaluating this claim?
2.People claim that you need to “liberate your
right brain” in order to be creative. What’s true
about this claim? What’s false about this claim?
E eBOOK DEMONSTRATIONS & ESSAYS
Go to http://digital.wwnorton.com/cognition7 for the online demonstrations and essays
relevant to this chapter. These can be found in the ebook.
Online Demonstrations
• Demonstration 2.1: Brain Anatomy
• Demonstration 2.2: The Speed of Neural
Online Applying Cognitive Psychology and the
Law Essays
• Cognitive Psychology and the Law: Improving
the Criminal Justice System
Transmission
• Demonstration 2.3: “Acuity” in the
Somatosensory System
COGNITION LABS
Go to http://digital.wwnorton.com/cognition7 to experience interactive online psychology
labs relevant to this chapter.
60
Learning about
the World
around Us
2
part
I
n setting after setting, you rely on your knowledge and beliefs. But where does
knowledge come from? The answer, usually, is experience, but this invites
another question: What is it that makes experience possible? Tackling this issue
will force us to examine mental processes that turn out to be surprisingly complex.
In Chapter 3, we’ll ask how visual perception operates — and, therefore, how
you manage to perceive the world around you. We’ll start with events in the
eyeball and then move to how you organize and interpret the visual information you receive. We’ll also consider the ways in which you can misinterpret this
information, so that you’re vulnerable to illusions.
Chapter 4 takes a further step and asks how you manage to recognize and
categorize the objects that you see. We’ll start with a simple case: how you
recognize printed letters. We’ll then turn to the recognition of more complex
(three-dimensional) objects.
In both chapters, we’ll discuss the active role that you play in shaping your
experience. We’ll see, for example, that your perceptual apparatus doesn’t just
“pick up” the information that’s available to you. You don’t, in other words, just
open your eyes and let the information “flow in.” Instead, we’ll discuss the ways in
which you supplement and interpret the information you receive. In Chapter 3, we’ll
see that this activity begins very early in the sequence of biological events that
support visual perception. In Chapter 4, these ideas will lead us to a mechanism
made up of very simple components, but shaped by a broad pattern of knowledge.
Chapter 5 then turns to the study of attention. As we’ll see, paying attention is a complex achievement involving many elements. We’ll discuss how the
mechanisms of attention can sometimes limit what people achieve — and so
part of what’s at stake in Chapter 5 is the question of what people ultimately
can or cannot accomplish, and whether there may be ways to escape these
apparent limits on human performance.
61
3
chapter
Visual Perception
what if…
You look around the world and instantly, effortlessly,
recognize the objects that surround you — words on this
page, objects in the room where you’re sitting, things you can view out
the window. Perception, in other words, seems fast, easy, and automatic.
But even so, there is complexity here, and your ability to perceive the
world depends on many separate and individually complicated processes.
Consider the disorder akinetopsia (Zeki, 1991). This condition is rare,
and much of what we know comes from a single patient — L.M. — who
developed this disorder because of a blood clot in her brain, at age 43.
L.M. was completely unable to perceive motion — even though other
aspects of her vision (e.g., her ability to recognize objects, to see color,
or to discern detail in a visual pattern) seemed normal.
Because of her akinetopsia, L.M. can detect that an object now is in a
position different from its position a moment ago, but she reports seeing “nothing in between.” As a way of capturing this experience, think
about what you see when you’re looking at really slow movement. If, for
example, you stare at the hour hand on a clock as it creeps around the
clock face, you cannot discern its motion. But you can easily see the
hand is now pointing, say, at the 4, and if you come back a while later,
you can see that it’s closer to the 5. In this way, you can infer motion from
the change in position, but you can’t perceive the motion. This is your
experience with very slow movement; L.M., suffering from akinetopsia,
has the same experience with all movement.
What’s it like to have this disorder? L.M. complained, as one concern,
that it was hard to cross the street because she couldn’t tell which of the
cars in view were moving and which ones were parked. (She eventually
learned to estimate the position and movement of traffic by listening
to cars’ sounds as they approached, even though she couldn’t see their
movement.)
Other problems caused by akinetopsia are more surprising. For
example, L.M. complained about difficulties in following conversations,
because she was essentially blind to the speaker’s lip movement or
changing facial expressions. She also felt insecure in social settings. If
more than two people were moving around in a room, she felt anxious
because “people were suddenly here or there, but [she had] not seen
them moving” (Zihl, von Cramon, & Mai, 1983, p. 315). Or, as a different
example: She had trouble in everyday activities like pouring a cup of
63
preview of chapter themes
•
•
e explore vision — humans’ dominant sensory modality.
W
We discuss the mechanisms through which the visual
system detects patterns in the incoming light, but we also
showcase the activity of the visual system in interpreting
and shaping the incoming information.
e also highlight the ways in which perception of
W
one aspect of the input is shaped by perception of other
aspects — so that the detection of simple features depends
on how the overall form is organized, and the perception of
size depends on the perceived distance of the target object.
•
e emphasize that the interpretation of the visual input
W
is usually accurate — but the same mechanisms can lead to
illusions, and the study of those illusions can often illuminate the processes through which perception functions.
coffee. She couldn’t see the fluid level’s gradual rise as she poured, so
she didn’t know when to stop pouring. For her, “the fluid appeared to be
frozen, like a glacier” (Zihl et al., 1983, p. 315; also Schenk, Ellison, Rice, &
Milner, 2005; Zihl, von Cramon, Mai, & Schmid, 1991).
We will have more to say about cases of disordered perception later in
the chapter. For now, though, let’s note the specificity of this disorder — a
disruption of movement perception, with other aspects of perception
still intact. Let’s also highlight the important point that each of us is, in
countless ways, dependent on our perceptual contact with the world.
That point demands that we ask: What makes this perception possible?
The Visual System
You receive information about the world through various sensory modalities:
You hear the sound of the approaching train, you smell the freshly baked
bread, you feel the tap on your shoulder. Researchers have made impressive
progress in studying all of these modalities, and students interested in, say,
hearing or the sense of smell will find a course in (or a book about) sensation
and perception to be fascinating.
There’s no question, though, that for humans vision is the dominant sense.
This is reflected in how much brain area is devoted to vision compared to
any of the other senses. It’s also reflected in many aspects of our behavior. For
example, if visual information conflicts with information received from other
senses, you usually place your trust in vision. This is the basis for ventriloquism,
in which you see the dummy’s mouth moving while the sounds themselves are
coming from the dummy’s master. Vision wins out in this contest, and so you
experience the illusion that the voice is coming from the dummy.
The Photoreceptors
How does vision operate? The process begins, of course, with light. Light is
produced by many objects in our surroundings — the sun, lamps, candles — and
then reflects off other objects. In most cases, it’s this reflected light — reflected
64 •
C H A P T E R T H R E E Visual Perception
from this book page or from a friend’s face — that launches the processes of
visual perception. Some of this light hits the front surface of the eyeball, passes
through the cornea and the lens, and then hits the retina, the light-sensitive
tissue that lines the back of the eyeball (see Figure 3.1). The cornea and lens
focus the incoming light, just as a camera lens might, so that a sharp image
is cast onto the retina. Adjustments in this process can take place because the
lens is surrounded by a band of muscle. When the muscle tightens, the lens
bulges somewhat, creating the proper shape for focusing the images cast by
nearby objects. When the muscle relaxes, the lens returns to a flatter shape,
allowing the proper focus for objects farther away.
On the retina, there are two types of photoreceptors — specialized neural
cells that respond directly to the incoming light. One type, the rods, are
sensitive to very low levels of light and so play an essential role whenever you’re moving around in semidarkness or trying to view a fairly dim
FIGURE 3.1
THE HUMAN EYE
Retina
Fovea
Pupil
Cornea
Lens
Iris
Optic nerve
(to brain)
Light enters the eye through the cornea, and the cornea and lens refract the
light rays to produce a sharply focused image on the retina. The iris can open or
close to control the amount of light that reaches the retina. The retina is made
up of three main layers: the rods and cones, which are the photoreceptors; the
bipolar cells; and the ganglion cells, whose axons make up the optic nerve.
The Visual System
•
65
RODS AND CONES
Cone
Rod
A
Thousands of photoreceptors
per square millimeter
FIGURE 3.2
Blind spot
180
Cones
Rods
140
Fovea
100
60
20
0
60
40
20
0
20
40
60
Distance on retina from fovea (degrees)
B
(Panel A) Rods and cones are the light-sensitive cells at the back of the retina that launch the neural process of
vision. In this (colorized) photo, cones appear green; rods appear brown. (Panel B) Distribution of photoreceptors.
Cones are most frequent at the fovea, and the number of cones drops off sharply as we move away from the fovea.
In contrast, there are no rods at all on the fovea. There are neither rods nor cones at the retina’s blind spot—the
position at which the neural fibers that make up the optic nerve exit the eyeball. Because this position is filled with
these fibers, there’s no space for any rods or cones.
stimulus. But the rods are also color-blind: They can distinguish different
intensities of light (and in that way contribute to your perception of brightness), but they provide no means of discriminating one hue from another
(see Figure 3.2).
Cones, in contrast, are less sensitive than rods and so need more incoming light to operate at all. But cones are sensitive to color differences. More
precisely, there are three different types of cones, each having its own pattern
of sensitivities to different wavelengths (see Figure 3.3). You perceive color,
therefore, by comparing the outputs from these three cone types. Strong
firing from only the cones that prefer short wavelengths, for example,
accompanied by weak (or no) firing from the other cone types, signals
purple. Blue is signaled by equally strong firing from the cones that prefer
short wavelengths and those that prefer medium wavelengths, with only
modest firing by cones that prefer long wavelengths. And so on, with other
patterns of firing, across the three cone types, corresponding to different
perceived hues.
Cones have another function: They enable you to discern fine detail. The
ability to see fine detail is referred to as acuity, and acuity is much higher for
66 •
C H A P T E R T H R E E Visual Perception
FIGURE 3.3
WAVELENGTHS OF LIGHT
Wavelength
Pressure
Amplitude
Time
A
White light
Prism
400
500
600
700 Nanometers
Visible light
Gamma
rays
X-rays
Ultraviolet
rays
Infrared
rays
Radar
Broadcast
radio
B
The physics of light are complex, but for many purposes light can be thought of as a wave (Panel A), and the
shape of the wave can be described in terms of its amplitude and its wavelength (i.e., the distance from “crest”
to “crest”). The wavelengths our visual system can sense are only a tiny part of the broader electromagnetic
spectrum (Panel B). Light with a wavelength longer than 750 nanometers is invisible to us, although we feel these
longer infrared waves as heat. Ultraviolet light, which has a wavelength shorter than 360 nanometers, is also
invisible to us. That leaves the narrow band of wavelengths between 750 and 360 nanometers — the so-called
visible spectrum. Within this spectrum, we usually see wavelengths close to 400 nanometers as violet, those
close to 700 nanometers as red, and those in between as the rest of the colors in the rainbow.
the cones than it is for the rods. This explains why you point your eyes toward
a target whenever you want to perceive it in detail. What you’re actually
doing is positioning your eyes so that the image of the target falls onto the
fovea, the very center of the retina. Here, cones far outnumber rods (and, in
fact, the center of the fovea has no rods at all). As a result, this is the region
of the retina with the greatest acuity.
The Visual System
•
67
In portions of the retina more distant from the fovea (i.e., portions of
the retina in the so-called visual periphery), the rods predominate; well out
into the periphery, there are no cones at all. This distribution of photo­
receptors explains why you’re better able to see very dim lights out of the
corner of your eyes. Psychologists have understood this point for at least a
century, but the key observation here has a much longer history. Sailors and
astronomers have known for hundreds of years that when looking at a
barely visible star, it’s best not to look directly at the star’s location. By
looking slightly away from the star, they ensured that the star’s image
would fall outside of the fovea and onto a region of the retina dense with
the more light-sensitive rods.
Lateral Inhibition
Rods and cones do not report directly to the cortex. Instead, the photoreceptors stimulate bipolar cells, which in turn excite ganglion cells. The
ganglion cells are spread uniformly across the entire retina, but all of their
axons converge to form the bundle of nerve fibers that we call the optic
nerve. This is the nerve tract that leaves the eyeball and carries information
to various sites in the brain. The information is sent first to a way station
in the thalamus called the lateral geniculate nucleus (LGN); from there,
information is transmitted to the primary projection area for vision, in the
occipital lobe.
Let’s be clear, though, that the optic nerve is not just a cable that conducts
signals from one site to another. Instead, the cells that link retina to brain are
already analyzing the visual input. One example lies in the phenomenon of
lateral inhibition, a pattern in which cells, when stimulated, inhibit the activity of neighboring cells. To see why this is important, consider two cells, each
receiving stimulation from a brightly lit area (see Figure 3.4). One cell (Cell B
in the figure) is receiving its stimulation from the middle of the lit area. It is
intensely stimulated, but so are its neighbors (including Cell A and Cell C).
As a result, all of these cells are active, and therefore each one is trying to inhibit its neighbors. The upshot is that the activity level of Cell B is increased
by the stimulation but decreased by the lateral inhibition it’s receiving from
Cells A and C. This combination leads to only a moderate level of activity
in Cell B.
In contrast, another cell (Cell C in the figure) is receiving its stimulation from the edge of the lit area. It is intensely stimulated, and so are its
neighbors on one side. Therefore, this cell will receive inhibition from one
side but not from the other (in the figure: inhibition from Cell B but
not from Cell D), so it will be less inhibited than Cell B (which is receiving
inhibition from both sides). Thus, Cells B and C initially receive the same
input, but C is less inhibited than B and so will end up firing more strongly
than B.
68 •
C H A P T E R T H R E E Visual Perception
FIGURE 3.4
LATERAL INHIBITION
Bright physical stimulus
Gray physical stimulus
Intense stimulation
Moderate stimulation
Cell
A
Cell
B
Cell
C
Cell
D
Cell
E
Cell
F
To brain
80 spikes
per second
To brain
90 spikes
per second
To brain
90 spikes
per second
Sideways
connection
providing
inhibition
To brain
1,000 spikes
per second
To brain
1,000 spikes
per second
To brain
1,100 spikes
per second
Stimulus as perceived
Stimulus as perceived
Cell B receives strong inhibition from all its neighbors, because its neighbors are intensely stimulated.
Cell C, in contrast, receives inhibition only from one side (because its neighbor on the other side, Cell D, is
only moderately stimulated). As a result, Cells B and C start with the same input, but Cell C, receiving less
inhibition, sends a stronger signal to the brain, emphasizing the edge in the stimulus. The same logic applies
to Cells D and E, and it explains why Cell D sends a weaker signal to the brain. Note, by the way, that the
spikes per second numbers, shown in the figure, are hypothetical and intended only to illustrate lateral
inhibition’s effects.
Notice that the pattern of lateral inhibition highlights a surface’s edges,
because the response of cells detecting the edge of the surface (such as Cell C)
will be stronger than that of cells detecting the middle of the surface (such as
Cell B). For that matter, by increasing the response by Cell C and decreasing
the response by Cell D, lateral inhibition actually exaggerates the contrast at
the edge — a process called edge enhancement. This process is of enormous
The Visual System
•
69
FIGURE 3.5
MACH BANDS
Edge enhancement, produced by lateral inhibition, helps us to perceive
the outline that defines an object’s shape. But the same process can
produce illusions — including the Mach bands. Each vertical strip in this figure is of uniform light intensity, but the strips don’t appear uniform. For
each strip, contrast makes the left edge (next to its darker neighbor) look
brighter than the rest, while the right edge (next to its lighter neighbor)
looks darker. To see that the differences are illusions, try placing a thin
object (such as a toothpick or a straightened paper clip) on top of the
boundary between strips. With the strips separated in this manner, the
illusion disappears.
TEST YOURSELF
1. What are the differences between rods
and cones? What traits
do these cells share?
2. What is lateral inhibition?
How does it contribute
to edge perception?
importance, because it’s obviously highlighting the information that defines an
object’s shape — information essential for figuring out what the object is. And
let’s emphasize that this edge enhancement occurs at a very early stage of
the visual processing. In other words, the information sent to the brain isn’t
a mere copy of the incoming stimulation; instead, the steps of interpretation
and analysis begin immediately, in the eyeball. (For a demonstration of an
illusion caused by this edge enhancement — the so-called Mach bands — see
Figure 3.5.)
Visual Coding
In Chapter 2, we introduced the idea of coding in the nervous system. This
term refers to the relationship between activity in the nervous system and the
stimulus (or idea or operation) that is somehow represented by that activity. In the study of perception, we can ask: What’s the code through which
neurons (or groups of neurons) manage to represent the shapes, colors, sizes,
and movements that you perceive?
70 •
C H A P T E R T H R E E Visual Perception
Single Neurons and Single-Cell Recording
Part of what we know about the visual system — actually, part of what we
know about the entire brain — comes from a technique called single-cell
recording. As the name implies, this is a procedure through which investigators can record, moment by moment, the pattern of electrical changes within
a single neuron.
We mentioned in Chapter 2 that when a neuron fires, each response is
the same size; this is the all-or-none law. But neurons can vary in how often
they fire, and when investigators record the activity of a single neuron, what
they’re usually interested in is the cell’s firing rate, measured in “spikes per
second.” The investigator can then vary the circumstances (either in the
external world or elsewhere in the nervous system) in order to learn what
makes the cell fire more and what makes it fire less. In this way, we can
figure out what job the neuron does within the broad context of the entire
nervous system.
The technique of single-cell recording has been used with enormous success in the study of vision. In a typical procedure, the animal being studied
is first immobilized. Then, electrodes are placed just outside a neuron in the
animal’s optic nerve or brain. Next, a computer screen is placed in front of
the animal’s eyes, and various patterns are flashed on the screen: circles, lines
at various angles, or squares of various sizes at various positions. Researchers
can then ask: Which patterns cause that neuron to fire? To what visual inputs
does that cell respond?
By analogy, we know that a smoke detector is a smoke detector because
it “fires” (i.e., makes noise) when smoke is on the scene. We know that a
motion detector is a motion detector because it “fires” when something
moves nearby. But what kind of detector is a given neuron? Is it responsive
to any light in any position within the field of view? In that case, we might
call it a “light detector.” Or is it perhaps responsive only to certain shapes
at certain positions (and therefore is a “shape detector”)? With this logic,
we can map out precisely what the cell responds to — what kind of detector
it is. More formally, this procedure allows us to define the cell’s receptive
field — that is, the size and shape of the area in the visual world to which
that cell responds.
Multiple Types of Receptive Fields
In 1981, the neurophysiologists David Hubel and Torsten Wiesel were
awarded the Nobel Prize for their exploration of the mammalian visual
system (e.g., Hubel & Wiesel, 1959, 1968). They documented the existence
of specialized neurons within the brain, each of which has a different type
of receptive field, a different kind of visual trigger. For example, some neurons seem to function as “dot detectors.” These cells fire at their maximum
TORSTEN WIESEL
AND DAVID HUBEL
Much of what we know about
the visual system is based on
the pioneering work done
by David Hubel and Torsten
Wiesel. This pair of researchers won the 1981 Nobel Prize
for their discoveries. (They
shared the Nobel with Roger
Sperry for his independent
research on the cerebral
hemispheres.)
Visual Coding
•
71
Receptive field
Center
Neural firing frequency
Time
A
Surround
Time
B
FIGURE 3.6
CELLS
CENTER-SURROUND
Some neurons in the visual system have receptive
fields with a “center-surround” organization. Panels A through D show the firing frequency for one
of those cells. (A) This graph shows the cell’s firing
rate when no stimulus is presented. (B) The cell’s
firing rate goes up when a stimulus is presented in
the middle of the cell’s receptive field. (C) In contrast, the cell’s firing rate goes down if a stimulus is
presented at the edge of the cell’s receptive field.
(D) If a stimulus is presented both to the center of
the receptive field and to the edge, the cell’s firing
rate does not change from its baseline level.
Time
C
Time
D
rate when light is presented in a small, roughly circular area in a specific
position within the field of view. Presentations of light just outside of this
area cause the cell to fire at less than its usual “resting” rate, so the input
must be precisely positioned to make this cell fire. Figure 3.6 depicts such
a receptive field.
These cells are often called center-surround cells, to mark the fact that light
presented to the central region of the receptive field has one influence, while light
presented to the surrounding ring has the opposite influence. If both the center and
the surround are strongly stimulated, the cell will fire neither more nor less than
usual. For this cell, a strong uniform stimulus is equivalent to no stimulus at all.
72 •
C H A P T E R T H R E E Visual Perception
FIGURE 3.7 ORIENTATION-SPECIFIC
VISUAL FIELDS
Some cells in the visual system fire only when the input contains a
line segment at a certain orientation. For example, one cell might
fire very little in response to a horizontal line, fire only occasionally
in response to a diagonal, and fire at its maximum rate only when
a vertical line is present. In this figure, the circles show the stimulus
that was presented. The right side shows records of neural firing.
Each vertical stroke represents a firing by the cell; the left–right
position reflects the passage of time.
( after hubel , 1963)
Other cells fire at their maximum only when a stimulus containing an
edge of just the right orientation appears within their receptive fields. These
cells, therefore, can be thought of as “edge detectors.” Some of these cells
fire at their maximum rate when a horizontal edge is presented; others,
when a vertical edge is in view; still others fire at their maximum to orientations in between horizontal and vertical. Note, though, that in each case,
these orientations merely define the cells’ “preference,” because these cells
are not oblivious to edges of other orientations. If a cell’s preference is for,
say, horizontal edges, then the cell will still respond to other orientations —
but less strongly than it does for horizontals. Specifically, the farther the
edge is from the cell’s preferred orientation, the weaker the firing will
be, and edges sharply different from the cell’s preferred orientation (e.g.,
a vertical edge for a cell that prefers horizontal) will elicit virtually no
response (see Figure 3.7).
Other cells, elsewhere in the visual cortex, have receptive fields that are
more specific. Some cells fire maximally only if an angle of a particular size
appears in their receptive fields; others fire maximally in response to corners
Visual Coding
•
73
and notches. Still other cells appear to be “movement detectors” and fire
strongly if a stimulus moves, say, from right to left across the cell’s receptive
field. Other cells favor left-to-right movement, and so on through the various
possible directions of movement.
Parallel Processing in the Visual System
This proliferation of cell types highlights another important principle —
namely, that the visual system relies on a “divide and conquer” strategy, with
different types of cells, located in different areas of the cortex, each specializing
in a particular kind of analysis. This pattern is plainly evident in Area V1, the
site on the occipital lobe where axons from the LGN first reach the cortex
(see Figure 3.8). In this brain area, some cells fire to (say) horizontals in this
position in the visual world, others to horizontals in that position, others to
verticals in specific positions, and so on. The full ensemble of cells in this area
FIGURE 3.8 AREA V1 IN
THE HUMAN BRAIN
Area V1 is the site on the occipital lobe where axons from the LGN
first reach the cortex. The top panel
shows the brain as if sliced vertically down the middle, revealing
the “inside” surface of the brain’s
right hemisphere. The bottom panel
shows the left hemisphere of the
brain viewed from the side. As the
two panels show, most of Area V1 is
located on the cortical surface bet­
ween the two cerebral hemispheres.
74 •
C H A P T E R T H R E E Visual Perception
Area V1, primary visual
projection area
provides a detector for every possible stimulus, making certain that no matter
what the input is or where it’s located, some cell will respond to it.
The pattern of specialization is also evident when we consider other
brain areas. Figure 3.9, for example, reflects one summary of the brain areas
known to be involved in vision. The details of the figure aren’t crucial, but it
is noteworthy that some of these areas (V1, V2, V3, V4, PO, and MT) are in
the occipital cortex; other areas are in the parietal cortex; others are in the
temporal cortex. (We’ll have more to say in a moment about these areas outside of the occipital cortex.) Most important, each area seems to have its own
function. Neurons in Area MT, for example, are acutely sensitive to direction
and speed of movement. (This area is the brain region that has suffered damage in cases involving akinetopsia.) Cells in Area V4 fire most strongly when
the input is of a certain color and a certain shape.
Let’s also emphasize that all of these specialized areas are active at the
same time, so that (for example) cells in Area MT are detecting movement in
the visual input at the same time that cells in Area V4 are detecting shapes.
In other words, the visual system relies on parallel processing — a system in
FIGURE 3.9
THE VISUAL PROCESSING PATHWAYS
VIP
Parietal
cortex
PO
MST
Occipital
cortex
7a
MT
LIP
Retina
LGN
V1
V3
V2
V4
TEO
Inferotemporal
cortex
TE
Each box in this figure refers to a specific location within the visual system. Notice that vision depends on many
brain sites, each performing a specialized type of analysis. Note also that the flow of information is complex, so
there’s no strict sequence of “this step” of analysis followed by “that step.” Instead, everything happens at once,
with a great deal of back-and-forth communication among the various elements.
Visual Coding
•
75
which many different steps (in this case, different kinds of analysis) are going
on simultaneously. (Parallel processing is usually contrasted with serial
processing, in which steps are carried out one at a time — i.e., in a series.)
One advantage of this simultaneous processing is speed: Brain areas
trying to discern the shape of the incoming stimulus don’t need to wait until
the motion analysis or the color analysis is complete. Instead, all of the analyses go forward immediately when the input appears before the eyes, with no
waiting time.
Another advantage of parallel processing is the possibility of mutual influence among multiple systems. To see why this matters, consider the fact that
sometimes your interpretation of an object’s motion depends on your understanding of the object’s three-dimensional shape. This suggests that it might
be best if the perception of shape happened first. That way, you could use
the results of this processing step as a guide to later analyses. In other cases,
though, the relationship between shape and motion is reversed. In these cases,
your interpretation of an object’s three-dimensional shape depends on your
understanding of its motion. To allow for this possibility, it might be best if
the perception of motion happened first, so that it could guide the subsequent
analysis of shape.
How does the brain deal with these contradictory demands? Parallel processing provides the answer. Since both sorts of analysis go on simultaneously,
each type of analysis can be informed by the other. Put differently, neither
the shape-analyzing system nor the motion-analyzing system gets priority.
Instead, the two systems work concurrently and “negotiate” a solution that
satisfies both systems (Van Essen & DeYoe, 1995).
Parallel processing is easy to document throughout the visual system. As
we’ve seen, the retina contains two types of specialized receptors (rods and
cones) each doing its own job (e.g., the rods detecting stimuli in the periphery
of your vision and stimuli presented at low light levels, and the cones detecting hues and detail at the center of your vision). Both types of receptors function at the same time — another case of parallel processing.
Likewise, within the optic nerve itself, there are two types of cells, P cells
and M cells. The P cells provide the main input for the LGN’s parvocellular
cells and appear to be specialized for spatial analysis and the detailed analysis
of form. M cells provide the input for the LGN’s magnocellular cells and are
specialized for the detection of motion and the perception of depth.1 And,
again, both of these systems are functioning at the same time — more parallel
processing.
Parallel processing remains in evidence when we move beyond the occipital cortex. As Figure 3.10 shows, some of the activation from the occipital
1. The names here refer to the relative sizes of the relevant cells: parvo derives from the Latin
word for “small,” and magno from the word for “large.” To remember the function of these two
types of cells, many students think of the P cells as specialized roughly for the perception of
pattern and M cells as specialized for the perception of motion. These descriptions are crude,
but they’re easy to remember.
76 •
C H A P T E R T H R E E Visual Perception
FIGURE 3.10
THE WHAT AND WHERE PATHWAYS
Parietal
lobe
Posterior
parietal
cortex
Temporal
lobe
Occipital
lobe
Inferotemporal
cortex
Information from the primary visual cortex at the back of the head is transmitted to the inferotemporal cortex (the so-called what system) and to the posterior parietal cortex (the where system). The term “inferotemporal” refers to
the lower part of the temporal lobe. The term “posterior parietal cortex” refers
to the rearmost portion of this cortex.
lobe is passed along to the cortex of the temporal lobe. This pathway, often
called the what system, plays a major role in the identification of visual
objects, telling you whether the object is a cat, an apple, or whatever. At the
same time, activation from the occipital lobe is also passed along a second
pathway, leading to the parietal cortex, in what is often called the where
system. This system seems to guide your action based on your perception of
where an object is located — above or below you, to your right or to your left.
(See Goodale & Milner, 2004; Humphreys & Riddoch, 2014; Ungerleider
& Haxby, 1994; Ungerleider & Mishkin, 1982. For some complications,
though, see Borst, Thompson, & Kosslyn, 2011; de Haan & Cowey, 2011.)
The contrasting roles of these two systems can be revealed in many ways,
including through studies of brain damage. Patients with lesions in the what
system show visual agnosia — an inability to recognize visually presented
objects, including such common things as a cup or a pencil. However, these
patients show little disorder in recognizing visual orientation or in reaching.
The reverse pattern occurs with patients who have suffered lesions in the
where system: They have difficulty in reaching, but no problem in object
identification (Damasio, Tranel, & Damasio, 1989; Farah, 1990; Goodale,
1995; Newcombe, Ratcliff, & Damasio, 1987).
Still other data echo this broad theme of parallel processing among separate systems. For example, we noted earlier that different brain areas are
Visual Coding
•
77
critical for the perception of color, motion, and form. If this is right, then
someone who has suffered damage in just one of these areas might show
problems in the perception of color but not the perception of motion or form,
or problems in the perception of motion but not the perception of form or
color. These predictions are correct. As we mentioned at the chapter’s start,
some patients suffer damage to the motion system and so develop akinetopsia (Zihl et al., 1983). For such patients, the world is described as a succession of static photographs. They’re unable to report the speed or direction of
a moving object; as one patient put it, “When I’m looking at the car first, it
seems far away. But then when I want to cross the road, suddenly the car is
very near” (Zihl et al., 1983, p. 315).
Other patients suffer a specific loss of color vision through damage to the
central nervous system, even though their perception of form and motion
remains normal (Damasio, 1985; Gazzaniga, Ivry, & Mangun, 2014; Meadows,
1974). To them, the entire world is clothed only in “dirty shades of gray.”2
Cases like these provide dramatic confirmation of the separateness of our
visual system’s various elements and the ways in which the visual system is
vulnerable to very specific forms of damage. (For further evidence with neurologically intact participants, see Bundesen, Kyllingsbaek, & Larsen, 2003.)
Putting the Pieces Back Together
Let’s emphasize once again, therefore, that even the simplest of our intellectual achievements depends on an array of different, highly specialized brain
areas all working together in parallel. This was evident in Chapter 2 in our
consideration of Capgras syndrome, and the same pattern has emerged in
our description of the visual system. Here, too, many brain areas must work
together: the what system and the where system, areas specialized for the detection of movement and areas specialized for the identification of simple forms.
We have identified the advantages that come from this division of labor
and the parallel processing it allows. But the division of labor also creates a
problem: If multiple brain areas contribute to an overall task, how is their
functioning coordinated? When you see an athlete make an astonishing jump,
the jump itself is registered by motion-sensitive neurons, but your recognition
of the athlete depends on shape-sensitive neurons. How are the pieces put
back together? When you reach for a coffee cup but stop midway because
you see that the cup is empty, the reach itself is guided by the where system;
the fact that the cup is empty is registered by the what system. How are these
two streams of processing coordinated?
Investigators refer to this broad issue as the binding problem — the
task of reuniting the various elements of a scene, elements that are initially
addressed by different systems in different parts of the brain. And obviously
2. This is different from ordinary color blindness, which is usually present from birth and results from
abnormalities that are outside the brain itself — for example, abnormalities in the photoreceptors.
78 •
C H A P T E R T H R E E Visual Perception
this problem is solved. What you perceive is not an unordered catalogue
of sensory elements. Instead, you perceive a coherent, integrated perceptual
world. Apparently, this is a case in which the various pieces of Humpty
Dumpty are reassembled to form an organized whole.
Visual Maps and Firing Synchrony
Look around you. Your visual system registers whiteness and blueness and
brownness; it also registers a small cylindrical shape (your coffee cup), a mediumsized rectangle (this book page), and a much larger rectangle (your desk). How
do you put these pieces together so that you see that it’s the coffee cup, and not
the book page, that’s blue; the desktop, and not the cup, that’s brown?
There is debate about how the visual system solves this problem, but we
can identify three elements that contribute to the solution. One element is
spatial position. The part of the brain registering the cup’s shape is separate
from the parts registering its color or its motion; nonetheless, these various
brain areas all have something in common. They each keep track of where
the target is — where the cylindrical shape was located, and where the blueness was; where the motion was detected, and where things were still. As a
result, the reassembling of these pieces can be done with reference to position. In essence, you can overlay the map of which forms are where on top
of the map of which colors are where to get the right colors with the right
forms, and likewise for the map showing which motion patterns are where.
Information about spatial position is, of course, useful for its own sake:
You have a compelling reason to care whether the tiger is close to you or
far away, or whether the bus is on your side of the street or the other. But in
addition, location information apparently provides a frame of reference used
to solve the binding problem. Given this double function, we shouldn’t be
surprised that spatial position is a major organizing theme in all the various
brain areas concerned with vision, with each area seeming to provide its own
map of the visual world.
Spatial position, however, is not the whole story. Evidence also suggests
that the brain uses special rhythms to identify which sensory elements belong
with which. Imagine two groups of neurons in the visual cortex. One group
of neurons fires maximally whenever a vertical line is in view; another group
fires maximally whenever a stimulus is in view moving from a high position
to a low one. Let’s also imagine that right now a vertical line is presented
and it is moving downward; as a result, both groups of neurons are firing
strongly. How does the brain encode the fact that these attributes are bound
together, different aspects of a single object? There is evidence that the visual
system marks this fact by means of neural synchrony: If the neurons detecting
a vertical line are firing in synchrony with those signaling movement, then
these attributes are registered as belonging to the same object. If they aren’t
in synchrony, then the features aren’t bound together (Buzsáki & Draguhn,
2004; Csibra, Davis, Spratling, & Johnson, 2000; Elliott & Müller, 2000;
Fries, Reynolds, Rorie, & Desimone, 2001).
Visual Coding
•
79
TEST YOURSELF
3.How do researchers
use single-cell recording to reveal a cell’s
receptive field?
4.What are the advantages of parallel processing in the visual
system? What are the
disadvantages?
5. How is firing synchrony
relevant to the solution of the binding
problem?
What causes this synchrony? How do the neurons become synchronized
in the first place? Here, another factor appears to be important: attention.
We’ll have more to say about attention in Chapter 5, but for now let’s note
that attention plays a key role in binding together the separate features of a
stimulus. (For a classic statement of this argument, see Treisman & Gelade,
1980; Treisman, Sykes, & Gelade, 1977. For more recent views, see Quinlan,
2003; Rensink, 2012; and also Chapter 5.)
Evidence for attention’s role comes from many sources, including the fact
that when we overload someone’s attention, she is likely to make conjunction
errors. This means that she’s likely to correctly detect the features present in
a visual display, but then to make mistakes about how the features are bound
together (or conjoined). Thus, for example, someone shown a blue H and a
red T might report seeing a blue T and a red H — an error in binding.
Similarly, individuals who suffer from severe attention deficits (because
of brain damage in the parietal cortex) are particularly impaired in tasks
that require them to judge how features are conjoined to form complex
objects (e.g., Robertson, Treisman, Friedman-Hill, & Grabowecky, 1997).
Finally, studies suggest that synchronized neural firing occurs in an animal’s
brain when the animal is attending to a specific stimulus but does not occur
in neurons activated by an unattended stimulus (e.g., Buschman & Miller,
2007; Saalmann, Pigarev, & Vidyasagar, 2007; Womelsdorf et al., 2007). All
of these results point toward the claim that attention is crucial for the binding
problem and, moreover, that attention is linked to the neural synchrony that
seems to unite a stimulus’s features.
Notice, then, that there are several ways in which information is represented
in the brain. In Chapter 2, we noted that the brain uses different chemical signals
(i.e., different neurotransmitters) to transmit different types of information.
We now see that there is information reflected in which cells are firing, how often
they are firing, whether the cells are firing in synchrony with other cells, and the
rhythm in which they are firing. Plainly, this is a system of considerable complexity!
Form Perception
So far in this chapter, we’ve been discussing how visual perception begins:
with the detection of simple attributes in the stimulus — its color, its motion,
and its catalogue of features. But this detection is just the start of the process,
because the visual system still has to assemble these features into recognizable
wholes. We’ve mentioned the binding problem as part of this “assembly” — but
binding isn’t the whole story. This point is reflected in the fact that our perception of the visual world is organized in ways that the stimulus input is
not — a point documented early in the 20th century by a group called the
“Gestalt psychologists.”3 The Gestaltists argued that the organization is
3. Gestalt is the German word for “shape” or “form.” The Gestalt psychology movement was
committed to the view that theories about perception and thought need to emphasize the
organization of patterns, not just focus on a pattern’s elements.
80 •
C H A P T E R T H R E E Visual Perception
contributed by the perceiver; this is why, they claimed, the perceptual whole
is often different from the sum of its parts. Some years later, Jerome Bruner
(1973) voiced related claims and coined the phrase “beyond the information
given” to describe some of the ways our perception of a stimulus differs from
(and goes beyond) the stimulus itself.
For example, consider the form shown in the top of Figure 3.11: the
Necker cube. This drawing is an example of a reversible (or ambiguous)
figure — so-called because people perceive it first one way and then another.
Specifically, this form can be perceived as a drawing of a cube viewed from
above (in which case it’s similar to the cube marked A in the figure); it can also
be perceived as a cube viewed from below (in which case it’s similar to the
cube marked B). Let’s be clear, though, that this isn’t an “illusion,” because neither of these interpretations is “wrong,” and the drawing itself (and, therefore,
the information reaching your eyes) is fully compatible with either interpretation. Put differently, the drawing shown in Figure 3.11 is entirely neutral
with regard to the shape’s configuration in depth; the lines on the page don’t
specify which is the “proper” interpretation. Your perception of the cube, however, is not neutral. Instead, you perceive the cube as having one configuration
or the other — similar either to Cube A or to Cube B. Your perception goes
beyond the information given in the drawing, by specifying an arrangement
in depth.
FIGURE 3.11
A
THE NECKER CUBE
B
The top cube can be perceived as if viewed from above (in which case it is a
transparent version of Cube A) or as if viewed from below (in which case it is
a transparent version of Cube B).
Form Perception
•
81
FIGURE 3.12
A
AMBIGUOUS FIGURES
B
Some stimuli easily lend themselves to reinterpretation. The figure in Panel A,
for example, is perceived by many to be a white vase or candlestick on a
black background; others see it as two black faces shown in profile. A similar
bistable form is visible in the Canadian flag (Panel B).
The same point can be made for many other stimuli. Figure 3.12A (after
Rubin, 1915, 1921) can be perceived either as a vase centered in the picture
or as two profiles facing each other. The drawing by itself is compatible
with either of these perceptions, and so, once again, the drawing is neutral
with regard to perceptual organization. In particular, it is neutral with regard to figure/ground organization, the determination of what is the figure
(the depicted object, displayed against a background) and what is the
ground. Your perception of this drawing, however, isn’t neutral about this
point. Instead, your perception somehow specifies that you’re looking at
the vase and not the profiles, or that you’re looking at the profiles and not
the vase.
Figure/ground ambiguity is also detectable in the Canadian flag (Figure 3.12B).
Since 1965, the centerpiece of Canada’s flag has been a red maple leaf. Many
observers, however, note that a different organization is possible, at least for
part of the flag. On their view, the flag depicts two profiles, shown in white
against a red backdrop. Each profile has a large nose, an open mouth, and
a prominent brow ridge, and the profiles are looking downward, toward the
flag’s center.
In all these examples, then, your perception contains information — about
how the form is arranged in depth, or about which part of the form is figure
and which is ground — that is not contained within the stimulus itself. Apparently, this is information contributed by you, the perceiver.
82 •
C H A P T E R T H R E E Visual Perception
The Gestalt Principles
With figures like the Necker cube or the vase/profiles, your role in shaping
the perception seems undeniable. In fact, if you stare at either of these figures,
your perception flips back and forth — first you see the figure one way, then
another, then back to the first way. But the stimulus itself isn’t changing, and
so the information that’s reaching your eyes is constant. Any changes in perception, therefore, are caused by you and not by some change in the stimulus.
One might argue, though, that reversible figures are special — carefully
designed to support multiple interpretations. On this basis, perhaps you play
a smaller role when perceiving other, more “natural” stimuli.
This position is plausible — but wrong, because many stimuli (and not just
the reversible figures) are ambiguous and in need of interpretation. We often
don’t detect this ambiguity, but that’s because the interpretation happens so
quickly that we don’t notice it. Consider, for example, the scene shown in
Figure 3.13. It’s almost certain that you perceive segments B and E as being
FIGURE 3.13
HE ROLE OF INTERPRETATION IN
T
PERCEIVING AN ORDINARY SCENE
C
A
E
B
D
A
B
Consider the still life (Panel A) and an overlay designating five different
segments of the scene (Panel B). For this picture to be perceived correctly,
the perceptual system must first decide what goes with what — for example,
that Segment B and Segment E are different bits of the same object (even
though they’re separated by Segment D) and that Segment B and Segment A
are different objects (even though they’re adjacent and the same color).
Form Perception
•
83
united, forming a complete apple, but notice that this information isn’t provided by the stimulus; instead, it’s your interpretation. (If we simply go with the
information in the figure, it’s possible that segments B and E are parts of
entirely different fruits, with the “gap” between the two fruits hidden from view
by the banana.) It’s also likely that you perceive the banana as entirely bananashaped and therefore continuing downward out of your view, into the bowl,
where it eventually ends with the sort of point that’s normal for a banana. In
the same way, surely you perceive the horizontal stripes in the background as
continuous and merely hidden from view by the pitcher. (You’d be surprised if
we removed the pitcher and revealed a pitcher-shaped gap in the stripes.) But,
of course, the stimulus doesn’t in any way “guarantee” the banana’s shape or
the continuity of the stripes; these points are, again, just your interpretation.
Even with this ordinary scene, therefore, your perception goes “beyond
the information given” — and so the unity of the two apple slices and the
continuity of the stripes is “in the eye of the beholder,” not in the stimulus itself. Of course, you don’t feel like you’re “interpreting” this picture or
extrapolating beyond what’s on the page. But your role becomes clear the
moment we start cataloguing the differences between your perception and
the information that’s truly present in the photograph.
Let’s emphasize, though, that your interpretation of the stimulus isn’t careless or capricious. Instead, you’re guided by a few straightforward principles
that the Gestalt psychologists catalogued many years ago — and so they’re
routinely referred to as the Gestalt principles. For example, your perception
is guided by proximity and similarity: If, within the visual scene, you see
elements that are close to each other, or elements that resemble each other,
you assume these elements are parts of the same object (Figure 3.14). You
also tend to assume that contours are smooth, not jagged, and you avoid
FIGURE 3.14
Similarity
We tend to group
these dots into
columns rather
than rows,
grouping dots of
similar colors.
GESTALT PRINCIPLES OF ORGANIZATION
Proximity
We tend to
perceive groups,
linking dots that
are close together.
Good continuation
We tend to see a
continuous green
bar rather than two
smaller rectangles.
Closure
We tend to perceive
an intact triangle,
reflecting our bias
toward perceiving
closed figures rather
than incomplete ones.
Simplicity
We tend to interpret a form
in the simplest way possible.
We would see the form on
the left as two intersecting
rectangles (as shown on the
right) rather than as a single
12-sided irregular polygon.
As Figure 3.13 illustrated, your ordinary perception of the world requires you to make decisions about what goes
with what — which elements are part of the same object, and which elements belong to different objects. Your
decisions are guided by a few simple principles, catalogued many years ago by the Gestalt psychologists.
84 •
C H A P T E R T H R E E Visual Perception
interpretations that involve coincidences. (For a modern perspective on these
principles and Gestalt psychology in general, see Wagemans, Elder, Kubovy
et al., 2012; Wagemans, Feldman, Gephstein et al., 2010.)
These perceptual principles are quite straightforward, but they’re essential if
your perceptual apparatus is going to make sense of the often ambiguous, often
incomplete information provided by your senses. In addition, it’s worth mentioning that everyone’s perceptions are guided by the same principles, and that’s why
you generally perceive the world in the same way that other people do. Each of
us imposes our own interpretation on the perceptual input, but we all tend to
impose the same interpretation because we’re all governed by the same rules.
Organization and Features
We’ve now considered two broad topics — the detection of simple attributes
in the stimulus, and then the ways in which you organize those attributes. In
thinking about these topics, you might want to think about them as separate
steps. First, you collect information about the stimulus, so that you know
(for example) what corners or angles or curves are in view — the visual features contained within the input. Then, once you’ve gathered the “raw data,”
you interpret this information. That’s when you “go beyond the information
given” — deciding how the form is laid out in depth (as in Figure 3.11), deciding what is figure and what is ground (Figure 3.12A or B), and so on.
The idea, then, is that perception might be divided (roughly) into an
“information gathering” step followed by an “interpretation” step. This view,
however, is wrong, and, in fact, it’s easy to show that in many settings, your
interpretation of the input happens before you start cataloguing the input’s
basic features, not after. Consider Figure 3.15. Initially, these shapes seem to
FIGURE 3.15
A HIDDEN FIGURE
Initially, these dark shapes have no meaning, but after a moment the hidden
figure becomes clearly visible. Notice, therefore, that at the start the figure
seems not to contain the features needed to identify the various letters. Once
the figure is reorganized, with the white parts (not the dark parts) making up
the figure, the features are easily detected. Apparently, the analysis of features depends how the figure is first organized by the viewer.
Form Perception
•
85
have no meaning, but after a moment most people discover the word hidden
in the figure. That is, people find a way to reorganize the figure so that the
familiar letters come into view. But let’s be clear about what this means. At
the start, the form seems not to contain the features needed to identify the L,
the I, and so on. Once the form is reorganized, though, it does contain these
features, and the letters are immediately recognized. In other words, with one
organization, the features are absent; with another, they’re plainly present. It
would seem, then, that the features themselves depend on how the form is
organized by the viewer — and so the features are as much “in the eye of the
beholder” as they are in the figure itself.
As a different example, you have no difficulty reading the word printed
in Figure 3.16, although most of the features needed for this recognition are
absent. You easily “provide” the missing features, though, thanks to the fact
that you interpret the black marks in the figure as shadows cast by solid
letters. Given this interpretation and the extrapolation it involves, you can
easily “fill in” the missing features and read the word.
How should we think about all of this? On one hand, your perception of a
form surely has to start with the stimulus itself and must in some ways be governed by what’s in that stimulus. (After all, no matter how you try to interpret
Figure 3.16, it won’t look to you like a photograph of Queen Elizabeth — the
basic features of the queen are just not present, and your perception respects
this obvious fact.) This suggests that the features must be in place before an
interpretation is offered, because the features govern the interpretation. But,
on the other hand, Figures 3.15 and 3.16 suggest that the opposite is the case:
that the features you find in an input depend on how the figure is interpreted.
Therefore, it’s the interpretation, not the features, that must be first.
The solution to this puzzle, however, is easy, and builds on ideas that we’ve
already met: Many aspects of the brain’s functioning depend on parallel processing, with different brain areas all doing their work at the same time. In
addition, the various brain areas all influence one another, so that what’s
going on in one brain region is shaped by what’s going on elsewhere. In this
FIGURE 3.16
MISSING FEATURES
PERCEPTION
People have no trouble reading this word, even though most of the features
needed for recognition are absent from the stimulus. People easily “supply”
the missing features, illustrating once again that the analysis of features
depends on how the overall figure has been interpreted and organized.
86 •
C H A P T E R T H R E E Visual Perception
way, the brain areas that analyze a pattern’s basic features do their work at
the same time as the brain areas that analyze the pattern’s large-scale configuration, and these brain areas interact so that the perception of the features is
guided by the configuration, and analysis of the configuration is guided by
the features. In other words, neither type of processing “goes first.” Neither
has priority. Instead, they work together, with the result that the perception
that is achieved makes sense at both the large-scale and fine-grained levels.
Constancy
We’ve now seen many indications of the perceiver’s role in “going beyond the
information given” in the stimulus itself. This theme is also evident in another
aspect of perception: the achievement of perceptual constancy. This term
refers to the fact that we perceive the constant properties of objects in the world
(their sizes, shapes, and so on) even though the sensory information we receive
about these attributes changes whenever our viewing circumstances change.
To illustrate this point, consider the perception of size. If you happen to
be far away from the object you’re viewing, then the image cast onto your
retinas by that object will be relatively small. If you approach the object, then
the image size will increase. This change in image size is a simple consequence
of physics, but you’re not fooled by this variation. Instead, you manage to
achieve size constancy — you correctly perceive the sizes of objects despite the
changes in retinal-image size created by changes in viewing distance.
Similarly, if you view a door straight on, the retinal image will be rectangular; but if you view the same door from an angle, the retinal image will
have a different shape (see Figure 3.17). Still, you achieve shape constancy —
that is, you correctly perceive the shapes of objects despite changes in the
retinal image created by shifts in your viewing angle. You also achieve brightness constancy — you correctly perceive the brightness of objects whether
they’re illuminated by dim light or strong sun.
FIGURE 3.17
TEST YOURSELF
6. W
hat evidence tells us
that perception goes
beyond (includes more
information than) the
stimulus input?
7. W
hat are the Gestalt
principles, and how do
they influence visual
perception?
8.What evidence is there
that the perception
of an overall form
depends on the detection of features? What
evidence is there that
the detection of features depends on the
overall form?
SHAPE CONSTANCY
If you change your viewing angle, the shape of the retinal image cast by a target changes. In this figure, the
door viewed straight on casts a rectangular image on
your retina; the door viewed from an angle casts a trape­
zoidal image. Nonetheless, you generally achieve shape
constancy.
Constancy
•
87
Unconscious Inference
How do you achieve each of these forms of constancy? One hypothesis
focuses on relationships within the retinal image. In judging size, for example, you generally see objects against some background, and this can provide
a basis for comparison with the target object. To see how this works, imagine
that you’re looking at a dog sitting on the kitchen floor. Let’s say the dog is
half as tall as the nearby chair and hides eight of the kitchen’s floor tiles from
view. If you take several steps back from the dog, none of these relationships
change, even though the sizes of all the retinal images are reduced. Size constancy, therefore, might be achieved by focusing not on the images themselves
but on these unchanging relationships (see Figure 3.18).
Relationships do contribute to size constancy, and that’s why you’re better
able to judge size when comparison objects are in view or when the target
you’re judging sits on a surface that has a uniform visual texture (like the
floor tiles in the example). But these relationships don’t tell the whole story.
Size constancy is achieved even when the visual scene offers no basis for comparison (if, for example, the object to be judged is the only object in view),
provided that other cues signal the distance of the target object (Harvey &
Leibowitz, 1967; Holway & Boring, 1947).
How does your visual system use this distance information? More than
a century ago, the German physicist Hermann von Helmholtz developed an
influential hypothesis regarding this question. Helmholtz started with the
fact that there’s a simple inverse relationship between distance and retinal image size: If an object doubles its distance from the viewer, the size of
its image is reduced by half. If an object triples its distance, the size of its
FIGURE 3.18 AN INVARIANT
RELATIONSHIP THAT
PROVIDES INFORMATION
ABOUT SIZE
One proposal is that you achieve size
constancy by focusing on relation­
ships in the visual scene. For example,
the dog sitting nearby on the kitchen
floor (Panel A) is half as tall as the
chair and hides eight of the kitchen’s
floor tiles from view. If you take several
steps back from the dog (Panel B),
none of these relationships change,
even though the sizes of all the retinal images are reduced. By focusing
on the relationships, then, you can see
that the dog’s size hasn’t changed.
88 •
A
C H A P T E R T H R E E Visual Perception
B
image is reduced to a third of its initial size. This relationship is guaranteed
to hold true because of the principles of optics, and the relationship makes
it possible for perceivers to achieve size constancy by means of a simple
calculation. Of course, Helmholtz knew that we don’t run through a conscious calculation every time we perceive an object’s size, but he believed
we’re calculating nonetheless — and so he referred to the process as an
unconscious inference (Helmholtz, 1909).
What is the calculation that enables someone to perceive size correctly?
It’s multiplication: the size of the image on the retina, multiplied by the distance between you and the object. (We’ll have more to say about how you
know this distance in a later section.) As an example, imagine an object that,
at a distance of 10 ft, casts an image on the retina that’s 4 mm across. Because
of straightforward principles of optics, the same object, at a distance of 20 ft,
casts an image of 2 mm. In both cases, the product — 10 3 4 or 20 3 2 — is
the same. If, therefore, your size estimate depends on that product, your size
estimate won’t be thrown off by viewing distance — and that’s exactly what
we want (see Figure 3.19).
FIGURE 3.19
THE RELATIONSHIP BETWEEN IMAGE SIZE AND DISTANCE
Closer objects cast larger retinal images
d
Retinal
image
Farther objects cast smaller retinal images
2d
Retinal
image
If you view an object from a greater distance, the object casts a smaller image on your retina. Nonetheless, you
generally achieve size constancy — perceiving the object’s actual size. Helmholtz proposed that you achieve constancy through an unconscious inference—essentially multiplying the image size by the distance.
Constancy
•
89
What’s the evidence that size constancy does depend on this sort of inference? In many experiments, researchers have shown participants an object
and, without changing the object’s retinal image, have changed the apparent
distance of the object. (There are many ways to do this — lenses that change
how the eye has to focus to bring the object into sharp view, or mirrors that
change how the two eyes have to angle inward so that the object’s image
is centered on both foveas.) If people are — as Helmholtz proposed — using
distance information to judge size, then these manipulations should affect
size perception. Any manipulation that makes an object seem farther away
(without changing retinal image size) should make that object seem bigger
(because, in essence, the perceiver would be “multiplying” by a larger number). Any manipulation that makes the object seem closer should make it
look smaller. And, in fact, these predictions are correct — a powerful confirmation that people do use distance to judge size.
A similar proposal explains how people achieve shape constancy. Here,
you take the slant of the surface into account and make appropriate
adjustments — again, an unconscious inference — in your interpretation of
the retinal image’s shape. Likewise for brightness constancy: Perceivers are
sensitive to how a surface is oriented relative to the available light sources,
and they take this information into account in estimating how much light
is reaching the surface. Then, they use this assessment of lighting to judge
the surface’s brightness (e.g., whether it’s black or gray or white). In all
these cases, therefore, it appears that the perceptual system does draw
some sort of unconscious inference, taking viewing circumstances into
account in a way that enables you to perceive the constant properties of
the visual world.
Illusions
This process of taking information into account — whether it’s distance (in
order to judge size), viewing angle (to judge shape), or illumination (to judge
brightness) — is crucial for achieving constancy. More than that, it’s another
indication that you don’t just “receive” visual information; instead, you interpret it. The interpretation is an essential part of your perception and generally helps you perceive the world correctly.
The role of the interpretation becomes especially clear, however, in circumstances in which you misinterpret the information available to you and end
up misperceiving the world. Consider the two tabletops shown in Figure 3.20.
The table on the left looks quite a bit longer and thinner than the one on the
right; a tablecloth that fits one table surely won’t fit the other. Objectively,
though, the parallelogram depicting the left tabletop is exactly the same shape
as the one depicting the right tabletop. If you were to cut out the shape on the
page depicting the left tabletop, rotate it, and slide it onto the right tabletop,
they’d be an exact match. (Not convinced? Just lay another piece of paper on
top of the page, trace the left tabletop, and then move your tracing onto the
right tabletop.)
90 •
C H A P T E R T H R E E Visual Perception
FIGURE 3.20
TWO TABLETOPS
These two tabletops seem to have very different shapes and sizes.
However, this contrast is an illusion — and the shapes drawn here
(the two parallelograms depicting the tabletops) are identical in
shape and size. The illusion is caused by the same mechanisms that,
in most circumstances, allow you to achieve constancy.
Why do people misperceive these shapes? The answer involves the normal
mechanisms of shape constancy. Cues to depth in this figure cause you to
perceive the figure as a drawing of three-dimensional objects, each viewed
from a particular angle. This leads you — quite automatically — to adjust for
the (apparent) viewing angles in order to perceive the two tabletops, and it’s
this adjustment that causes the illusion. Notice, then, that this illusion about
shape is caused by a misperception of depth: You misperceive the depth relationships in the drawing and then take this faulty information into account in
interpreting the shapes. (For a related illusion, see Figure 3.21.)
FIGURE 3.21 THE
MONSTER ILLUSION
The two monsters appear rather
different in size. But, again, this
is an illusion, because the two
drawings are exactly the same
size. The illusion is created by
the distance cues in the picture,
which make the monster on
the right appear to be farther
away. This (mis)perception of distance leads to a (mis)perception of size.
Constancy
•
91
FIGURE 3.22
A BRIGHTNESS ILLUSION
The central square (third row, third column) appears
much brighter than the square marked by the arrow.
Once again, though, this is an illusion. If you don’t
believe it, use your fingers or pieces of paper to cover
everything in the figure except for these two squares.
TEST YOURSELF
9.What does it mean to
say that size constancy may depend on an
unconscious inference? An inference
about what?
10. How do the ordinary
mechanisms of constancy lead to visual
illusions?
A different example is shown in Figure 3.22. It seems obvious to most
viewers that the center square in this checkerboard (third row, third column)
is a brighter shade than the square indicated by the arrow. But, in truth, the
shade of gray shown on the page is identical for these two squares. What has
happened here? The answer again involves the normal processes of perception.
First, the mechanisms of lateral inhibition (described earlier) play a role here
in producing a contrast effect: The central square in this figure is surrounded
by dark squares, and the contrast makes the central square look brighter. The
square marked at the edge of the checkerboard, however, is surrounded by
white squares; here, contrast makes the marked square look darker.
But, in addition, the visual system also detects that the central square is in
the shadow cast by the cylinder. Your vision compensates for this fact — again,
an example of unconscious inference that takes the shadow into account in
judging brightness — and therefore powerfully magnifies the illusion.
The Perception of Depth
In discussing constancy, we said that perceivers take distance, slant, and
illumination into account in judging size, shape, and brightness. But to do this,
they need to know what the distance is (how far away is the target object?),
what the viewing angle is (“Am I looking at the shape straight on or at an
angle?”), and what the illumination is. Otherwise, they’d have no way to take
these factors into account and, therefore, no way to achieve constancy.
Let’s pursue this issue by asking how people judge distance. We’ve just
said that distance perception is crucial for size constancy, but, of course,
information about where things are in your world is also valuable for its own
sake. If you want to walk down a hallway without bumping into obstacles,
you need to know which obstacles are close to you and which ones are far
off. If you wish to caress a loved one, you need to know where he or she is;
92 •
C H A P T E R T H R E E Visual Perception
otherwise, you’re likely to swat empty space when you reach out with your
caress or (worse) poke him or her in the eye. Plainly, then, you need to know
where objects in your world are located.
Binocular Cues
The perception of distance depends on various distance cues — features of the
stimulus that indicate an object’s position. One cue comes from the fact that
your eyes look out on the world from slightly different positions; as a result,
each eye has a slightly different view. This difference between the two eyes’
views is called binocular disparity, and it provides important information
about distance relationships in the world.
Binocular disparity can lead to the perception of depth even when no other
distance cues are present. For example, the bottom panels of Figure 3.23
show the views that each eye would receive while looking at a pair of nearby
objects. If we present each of these views to the appropriate eye (e.g., by drawing the views on two cards and placing one card in front of each eye), we can
obtain a striking impression of depth.
Monocular Cues
Binocular disparity is a powerful determinant of perceived depth. But we can
also perceive depth with one eye closed; plainly, then, there are also depth cues
that depend only on what each eye sees by itself. These are the monocular
distance cues.
B
A
B
A
A
B
Left eye's view
A B
Right eye's view
A
B
FIGURE 3.23
BINOCULAR DISPARITY
Your two eyes look out on the world from slightly different positions, and therefore they get slightly different views. The visual
system uses this difference in views as a cue to distance. This
figure shows what the left eye’s and right eye’s views would be
in looking at objects A and B.
The Perception of Depth
•
93
One monocular cue depends on the adjustment that the eye must make in
order to see the world clearly. We mentioned earlier that in each eye, muscles
adjust the shape of the lens to produce a sharply focused image on the retina.
The amount of adjustment depends on how far away the viewed object
is — there’s a lot of adjustment for nearby objects, less for those a few steps
away, and virtually no adjustment at all for objects more than a few meters
away. It turns out that perceivers are sensitive to the amount of adjustment
and use it as a cue indicating how far away the object is.
Other monocular cues have been exploited by artists for centuries to
create an impression of depth on a flat surface — that is, within a picture — and
that’s why these cues are called pictorial cues. In each case, these cues rely
on straightforward principles of physics. For example, imagine a situation
in which a man is trying to admire a sports car, but a mailbox is in the way
(see Figure 3.24A). In this case, the mailbox will inevitably block the view
simply because light can’t travel through an opaque object. This fact about
the physical world provides a cue you can use in judging distance. The cue is
known as interposition — the blocking of your view of one object by some
other object. In Figure 3.24B, interposition tells the man that the mailbox is
closer than the car.
In the same way, distant objects produce a smaller retinal image than do
nearby objects of the same size; this is a fact about optics. But this physical
fact again provides perceptual information you can use. In particular, it’s the
FIGURE 3.24
A
INTERPOSITION AS A DEPTH CUE
B
This man is looking at the sports car, but the mailbox blocks part of his view
(Panel A). Here’s how the scene looks from the man’s point of view (Panel B).
Because the mailbox blocks the view, the man gets a simple but powerful cue
that the mailbox must be closer to him than the sports car is.
94 •
C H A P T E R T H R E E Visual Perception
FIGURE 3.25
A
E FFECT OF CHANGES IN TEXTURE
GRADIENT
B
Changes in texture provide important information about spatial arrangements
in the world. Examples here show (Panel A) an upward tilt and (Panel B) a
sudden drop.
basis for the cue of linear perspective, the name for the pattern in which parallel lines seem to converge as they get farther and farther from the viewer.
A related cue is provided by texture gradients. Consider what meets your
eye when you look at cobblestones on a street or patterns of sand on a beach.
The retinal projection of the sand or cobblestones shows a pattern of continuous change in which the elements of the texture grow smaller and smaller
as they become more distant. This pattern of change by itself can reveal the
spatial layout of the relevant surfaces. If, in addition, there are discontinuities
in these textures, they can tell you even more about how the surfaces are laid
out (see Figure 3.25; Gibson, 1950, 1966).
The Perception of Depth through Motion
Whenever you move your head, the images projected by objects in your view
move across your retinas. For reasons of geometry, the projected images of
nearby objects move more than those of distant ones, and this pattern of
motion in the retinal images gives you another distance cue, called motion
parallax (Helmholtz, 1909).
A different cue relies on the fact that the pattern of stimulation across the
entire visual field changes as you move forward. This change in the visual
input — termed optic flow — provides another type of information about
depth and plays a large role in the coordination of bodily movements (Gibson,
1950, 1979).
The Perception of Depth
•
95
COGNITION
outside the lab
Virtual Reality
We obviously move around in a three-dimensional
(invented by a man whose son went on to be a
world. For centuries, though, people have been try-
Supreme Court Justice for thirty years!). This wooden
ing to create an illusion of 3-D with displays that are
device (Panel A in the figure below) allows the pre-
actually flat. Painters during the Renaissance, for ex-
sentation of a pair of pictures, one to each eye. The
ample, developed great skill in the use of the “picto-
two pictures show the same scene but viewed from
rial cues” (including visual perspective) to create a
slightly different vantage points, and these “ste-
sense of depth on a flat canvas. In the extreme, the
reoviews” produce a compelling sense of depth.
art technique of trompe l’oeil (French for “deceive
The same principle — and your capacity for “ste-
the eye”) could leave people truly puzzled about
reovision” — is used with the “virtual reality” (VR)
whether an object was painted or actually present.
accessory that works with many smartphones. The
Panel C in the figure on the following page shows
accessory, often made of cardboard, places a lens in
a modern version — created by a talented sidewalk
front of each eye so that you’ll be comfortable point-
artist.
ing your eyes straight ahead (as if you were looking
A different technique relies on binocular (“two-
at something far away), even though you’re actually
eyed”) vision. Consider the Holmes stereoscope
looking at an image just an inch or so away. With this
A
B
CLASSICAL USES OF BINOCULAR DISPARITY
Binocular disparity was the principle behind the stereoscope (Panel A), a device popular in the 19th century that
presented a slightly different photograph to each eye, creating a vivid sense of depth. The ViewMaster (Panel B),
a popular children’s toy, works in exactly the same way. The photos on the wheel are actually in pairs — and so, at
any rotation, the left eye views one photo in the pair (the one at 9 o’clock on the wheel) and the right eye views a
slightly different photo (the one at 3 o’clock), one that shows the same scene from a slightly different angle. Again,
the result is a powerful sense of depth.
96 •
C H A P T E R T H R E E Visual Perception
C
D
MODERN SIMULATIONS OF 3-D
Panel C shows a chalk drawing on a flat (and entirely undamaged) sidewalk. By manipulating pictorial cues, though,
the artist creates a compelling illusion of depth—with a car collapsed into a pit that in truth isn’t there at all.
Panel D shows one of the devices used to turn a smartphone into a “virtual reality” viewer.
setup, your phone displays two views of the same
viewers wear eyeglasses that contain corresponding
scene (one view to each eye), viewed from slightly
filters. The eyeglass filters “pass” light that’s polar-
different angles. Your eyes “fuse” these inputs into a
ized in a way that matches the filter, and block light
single image, but that doesn’t mean you ignore the
that’s polarized differently. As a result, each eye sees
differences between the inputs. Instead, your visual
only one of the projected movies — and, again, view-
system is solving the geometric puzzle posed by the
ers fuse the images but use the binocular disparity
two inputs. In other words, your brain manages to
to produce the experience of depth.
figure out how the scene must have been arranged
If you’ve enjoyed a 3-D movie or a smartphone
in order to produce these two different views, and
VR system, you’ve seen that the sense of depth is
it’s the end product of this computation that you
quite compelling. But these systems don’t work for
experience as a three-dimensional scene.
everyone. Some people have a strong pattern of
Three-D movies work the same way. There are
“ocular dominance,” which means that they rely on
actually two separate movies projected onto the
one eye far more than on the other. For these people,
theater’s screen. In some cases, the movies were
binocular disparity (which depends on combining
recorded from slightly different positions; in other
the inputs from both eyes) loses its force. However,
cases, the two perspectives were computer gen-
these people can still draw depth information from
erated. In either situation, the separate movies are
other (monocular or motion-based) cues, and so
projected through filters that polarize the light and
they can enjoy the same movies as anyone else.
The Perception of Depth
•
97
LINEAR PERSPECTIVE
AS A CUE FOR DEPTH
The Role of Redundancy
TEST YOURSELF
11. W
hat are the mono­
cular cues to distance?
12. Why is it helpful that
people rely on several
different cues in judging distance?
98 •
One might think that the various distance cues all end up providing the same
information — each one tells you which objects are close by and which ones
are distant. On that basis, it might be efficient for the visual system to focus
on just one or two cues and ignore the others. The fact is, however, that you
use all these cues, as well as several others we haven’t described (e.g., see
Figure 3.26).
Why is our visual system influenced by so many cues, especially since
these cues do, in fact, often provide redundant information? It’s because
different distance cues become important in different circumstances. For
example, binocular disparity is a powerful cue, but it’s informative only
when objects are relatively close by. (For targets farther than 30 ft away, the
two eyes receive virtually the same image.) Likewise, motion parallax tells
you a great deal about the spatial layout of your world, but only if you’re
moving. Texture gradients are informative only if there’s a suitably uniform
texture in view. So while these various cues are often redundant, each type
of cue can provide information when the others cannot. By being sensitive to them all, you’re able to judge distance in nearly any situation you
encounter. This turns out to be a consistent theme of perception — with
multiple cues to distance, multiple cues to illumination, multiple paths
through which to detect motion, and so on. The result is a system that
sometimes seems inelegant and inefficient, but it’s one that guarantees flexibility and versatility.
C H A P T E R T H R E E Visual Perception
FIGURE 3.26
M
ONOCULAR CLUES TO DEPTH:
LIGHT AND SHADOW
A
B
In this chapter, we’ve covered only a subset of the cues to distance that are
used by our visual system. Another cue is provided by the shadows “attached”
to an object. In Panel A, most viewers will say that the figure contains
six “bulges” in a smiley-face configuration (two eyes, a nose, a mouth).
In Panel B, the same figure has been turned upside-down. Now, the bulges
appear to be “dents,” and the other circles that appeared concave in the
Panel A view now look like bulges. The reason is the location of the shadows.
When the shadow is at the bottom, the object looks convex — a point that
makes sense because in our day-to-day lives light almost always comes from
above us, not below.
COGNITIVE PSYCHOLOGY AND EDUCATION
an “educated eye”
In the courtroom, eyewitnesses are often asked to describe what they saw
at a crime scene, and asked if they can identify the person who committed
the crime. Judges and juries generally rely on this testimony and accept the
witness’s report as an accurate description of how things unfolded. Judges
and juries are, however, especially likely to accept the witness’s report as
accurate if the witness is a police officer. In support of this position, some
attorneys argue that police officers have “educated eyes,” with the result
that police can (for example) recognize faces that they viewed only briefly or
at a considerable distance. In one trial, a police officer even claimed that
thanks to years of working a late-night shift, he’d improved his ability to
see in the dark.
Cognitive Psychology and Education
•
99
Related ideas arise in other settings. In Chapter 4, we’ll discuss programs
that teach you how to “speed-read,” but for now let’s just note that some of
these programs make a strong claim in their advertising: They claim that they
train your eyes so that you can “see more in a single glance.”
At one level, these claims are nonsensical. How much you can see “in
a single glance” is limited by your visual acuity, and acuity is limited by the
optical properties of the eyeball and the functional properties of the photoreceptors. To see more “in a single glance,” we’d need to give you a new cornea,
a new lens, and a new retina — and, of course, no speed-reading program
offers that sort of transplant surgery. Likewise, your ability to see in the dark
is constrained by the biological properties of the eye (including the structure
of the photoreceptors and the chemical principles that govern the photoreceptors’ response to light). No experience, and no training, is going to change
these properties.
At a different level, though, it is possible to have an “educated eye”— or,
more precisely, to be more observant and more discerning than other people.
For example, when looking at a complex, fast-moving crime scene, police
officers are more likely to focus their attention on details that will matter for the
investigation — and so will likely see (and remember) more of the perpetrator’s
actions (although, ironically, this means they’ll see less of what’s happening elsewhere in the scene). In the same way, referees and umpires in professional sports
know exactly what to focus on during a game. (Did the player’s knee touch the
ground before he fumbled the ball? Did the basketball player drag her pivot
foot or not?) As a result, they’ll see things that ordinary observers would miss.
These advantages (for police officers or for referees) may seem obvious,
but the advantages are closely tied to points raised in the chapter. You are
able to see detail only for visual inputs landing on your foveas; what lands
on your foveas depends on where exactly you’re pointing your eyes; and
movements of the eyes (pointing them first here and then there) turn out to be
relatively slow. As a result, knowledge about where to look has an immense
impact on what you’ll be able to see.
It’s also true that experience can help you to see certain patterns that
you’d otherwise miss. In some cases, the experience helps you to stop looking at a visual input on a feature-by-feature basis, but instead to take a more
“global” perspective so that you look at the pattern overall. Expert chess
players, for example, seem to perceive a chess board in terms of the patterns
in place (patterns indicating an upcoming attack, or patterns revealing the
opponent’s overall strategy), and this perspective helps them to plan their
own moves. (For more on chess experts, see Chapter 13.) Or, as a very different example, consider the dog experts who serve as judges at the Westminster
Kennel Club Dog Show. Evidence suggests that these experts are sensitive to
each dog’s overall form, and not just the shape of the front legs, the chest, the
ears, and so on, with the result that they can make more discerning assessments than an ordinary dog-lover could.
100 •
C H A P T E R T H R E E Visual Perception
KNOWING WHERE TO LOOK
Referees in football games know exactly where to look in order to pick up the
information they need in making their judgments. Did the player tap both feet
on the ground before going out of bounds? Did the player’s knee touch the
ground before he fumbled the ball? Referees seem to have an “educated eye,”
but, in reality, their advantage comes from how (and where) they focus their
attention.
Experience can also help you to see (or hear or feel) certain combinations that are especially important or informative. One prominent example
involves experienced firefighters who sometimes have an eerie ability to
judge when a floor is about to collapse — allowing these professionals to
evacuate a building in time, saving their lives and others’. What explains
this perception? The answer may be a combination of feeling an especially
high temperature and hearing relative quiet — a combination that signals a
ferocious fire burning underneath them, hidden under the floor that they’re
standing on.
In short, then, people can have “educated eyes” (or ears or noses or
palates). This “education” can’t change the basic biological properties of
your sense organs. But knowledge and experience can certainly help you
to see things that others overlook, to detect patterns that are largely invisible to other people, and to pick up on combinations that can — in some
settings — save your life.
Cognitive Psychology and Education
•
101
For more on this topic . . .
Biederman, I., & Shiffrar, M. M. (1987). Sexing day-old chicks: A case study and
expert systems analysis of a difficult perceptual learning task. Journal of
Experimental Psychology: Learning, Memory & Cognition, 13, 640–645.
Diamond, R., & Carey, S. (1986). Why faces are and are not special: An effect
of expertise. Journal of Experimental Psychology: General, 115, 107–117.
Klein, R. (2013). Seeing what others don’t. New York, NY: Public Affairs.
Vredeveldt, A., Knol, J. W., & van Kopen, P. J. (2017). Observing offenders:
Incident reports by surveillance detectives, uniformed police, and civilians.
Legal and Criminological Psychology, 22, 150–163.
102 •
C H A P T E R T H R E E Visual Perception
chapter review
SUMMARY
• One brain area that has been mapped in considerable detail is the visual system. This system takes
its main input from the rods and cones on the retina.
Then, information is sent via the optic nerve to the
brain. An important point is that cells in the optic
nerve do much more than transmit information;
they also begin the analysis of the visual input. This
is reflected in the phenomenon of lateral inhibition,
which leads to edge enhancement.
• Part of what we know about the brain comes
from single-cell recording, which can record the
electrical activity of an individual neuron. In the
visual system, this recording has allowed researchers to map the receptive fields for many cells. The
mapping has provided evidence for a high degree of
specialization among the various parts of the visual
system, with some parts specialized for the perception of motion, others for the perception of color,
and so on. The various areas function in parallel, and
this parallel processing allows great speed. It also
allows mutual influence among multiple systems.
• Parallel processing begins in the optic nerve and
continues throughout the visual system. For example, the what system (in the temporal lobe) appears
to be specialized for the identification of visual
objects; the where system (in the parietal lobe)
seems to identify where an object is located.
• The reliance on parallel processing creates
a problem of reuniting the various elements of a
scene so that these elements are perceived in an
integrated way. This is the binding problem. One
key in solving this problem lies in the fact that
different brain systems are organized in terms of
maps, so that spatial position can be used as a
framework for reuniting the separately analyzed
aspects of the visual scene.
• Visual perception requires more than the “pickup” of features. Those features must be organized
into wholes — a process apparently governed by the
so-called Gestalt principles. The visual system also
must interpret the input, a point that is especially
evident with reversible figures. Crucially, though,
these interpretive steps aren’t separate from, and
occurring after, the pickup of elementary features,
because the features themselves are shaped by the
perceiver’s organization of the input.
• The active nature of perception is also evident in
perceptual constancy. We achieve constancy through
a process of unconscious inference, taking one
aspect of the input (e.g., the distance to the target)
into account in interpreting another aspect (e.g., the
target’s size). This process is usually quite accurate,
but it can produce illusions.
• The perception of distance relies on many
cues — some dependent on binocular vision,
and some on monocular vision. The diversity of
cues lets us perceive distance in a wide range of
circumstances.
KEY TERMS
cornea (p. 65)
lens (p. 65)
retina (p. 65)
photoreceptors (p. 65)
rods (p. 65)
cones (p. 66)
acuity (p. 66)
fovea (p. 67)
103
bipolar cells (p. 68)
ganglion cells (p. 68)
optic nerve (p. 68)
lateral geniculate nucleus (LGN) (p. 68)
lateral inhibition (p. 68)
edge enhancement (p. 69)
Mach band (p. 70)
single-cell recording (p. 71)
receptive field (p. 71)
center-surround cells (p. 72)
Area V1 (p. 74)
parallel processing (p. 75)
serial processing (p. 76)
P cells (p. 76)
M cells (p. 76)
parvocellular cells (p. 76)
magnocellular cells (p. 76)
what system (p. 77)
where system (p. 77)
binding problem (p. 78)
neural synchrony (p. 79)
conjunction errors (p. 80)
Necker cube (p. 81)
reversible figure (p. 81)
figure/ground organization (p. 82)
Gestalt principles (p. 84)
visual features (p. 85)
perceptual constancy (p. 87)
size constancy (p. 87)
shape constancy (p. 87)
brightness constancy (p. 87)
unconscious inference (p. 89)
distance cues (p. 93)
binocular disparity (p. 93)
monocular distance cues (p. 93)
pictorial cues (p. 94)
interposition (p. 94)
linear perspective (p. 95)
motion parallax (p. 95)
optic flow (p. 95)
TEST YOURSELF AGAIN
1.What are the differences between rods and
cones? What traits do these cells share?
7.What are the Gestalt principles, and how do
they influence visual perception?
2.What is lateral inhibition? How does it
contribute to edge perception?
3.How do researchers use single-cell recording to
reveal a cell’s receptive field?
8.What evidence is there that the perception of
an overall form depends on the detection of
features? What evidence is there that the detection of features depends on the overall form?
4.What are the advantages of parallel
processing in the visual system? What are
the disadvantages?
9.What does it mean to say that size constancy
may depend on an unconscious inference? An
inference about what?
5.How is firing synchrony relevant to the solution
of the binding problem?
10.How do the ordinary mechanisms of constancy
lead to visual illusions?
6.What evidence tells us that perception goes
beyond (i.e., includes more information than)
the stimulus input?
11. What are the monocular cues to distance?
104
12.Why is it helpful that people rely on several
different cues in judging distance?
THINK ABOUT IT
1.The chapter emphasizes the active nature of
perception — and the idea that we don’t just
“pick up” information from the environment;
instead, we interpret and supplement that information. What examples of this pattern can you
think of — either from the chapter or from your
own experience?
2.Chapter 2 argued that the functioning of the
brain depends on the coordination of many
specialized operations. How does that claim,
about the brain in general, fit with the discussion of visual perception in this chapter?
E eBOOK DEMONSTRATIONS & ESSAYS
Go to http://digital.wwnorton.com/cognition7 for the online demonstrations and essays
relevant to this chapter. These can be found in the ebook.
Online Demonstrations
• Demonstration 3.1: Foveation
• Demonstration 3.2: Eye Movements
• Demonstration 3.3: The Blind Spot and the
• Demonstration 3.4: A Brightness Illusion
• Demonstration 3.5: A Size Illusion and a Motion
Illusion
Active Nature of Vision
COGNITION LABS
Go to http://digital.wwnorton.com/cognition7 to experience interactive online psychology
labs relevant to this chapter.
105
4
chapter
Recognizing
Objects
what if…
In Chapter 3, we discussed some of the steps involved
in visual perception — steps allowing you to see that
the object in front of you is, let’s say, brown, large, and moving. But
you don’t leave things there; you also recognize objects and can identify
what they are (perhaps: a UPS truck). This sort of recognition is usually
easy for you, so you have no difficulty in recognizing the vast array of
objects in your world — trucks, squirrels, shoes, frying pans, and more.
But, easy or not, recognition relies on processes that are surprisingly
sophisticated, and your life would be massively disrupted if you couldn’t
manage this (seemingly simple) achievement.
We mentioned in Chapter 2 that certain types of brain damage
produce a disorder called “agnosia.” In some cases, patients suffer from
apperceptive agnosia — they seem able to see an object’s shape and color
and position, but they can’t put these elements together to perceive
the entire object. For example, one patient — identified as D.F. — suffered
from brain damage in the sites shown in Figure 4.1. D.F. was asked
to copy drawings that were in plain view (Figure 4.2A). The resulting
attempts are shown in Figure 4.2B. The limit here is not some problem
in drawing ability. Figure 4.2C shows what happened when D.F. was
asked to draw various forms from memory. Plainly, D.F. can draw; the
problem instead is in her ability to see and assemble the various
elements that she sees.
Other patients suffer from associative agnosia. They can see but cannot link what they see to their basic visual knowledge. One remarkable
example comes from a case described by neurologist Oliver Sacks:
“What is this?” I asked, holding up a glove.
“May I examine it?” he asked, and, taking it from me, he proceeded to
examine it. “A continuous surface,” he announced at last, “infolded
in itself. It appears to have” — he hesitated — ”five outpouchings, if
this is the word.”
“Yes,” I said cautiously. “. . .Now tell me what it is.”
“A container of some sort?”
“Yes,” I said, “and what would it contain?”
“It would contain its contents!” said Dr. P., with a laugh. “There are
many possibilities. It could be a change purse, for example, for
coins of five sizes. It could . . .” (Sacks, 1985, p. 14)
107
preview of chapter themes
•
ecognition of visual inputs begins with features, but it’s not
R
just the features that matter. How easily people recognize a
pattern also depends on how frequently or recently they have
viewed the pattern and on whether the pattern is well formed
(such as letter sequences with “normal” spelling patterns).
•
e explain these findings in terms of a feature net — a netW
work of detectors, each of which is “primed” according to
how often or how recently it has fired. The network relies
on distributed knowledge to make inferences, and this
process gives up some accuracy in order to gain efficiency.
FIGURE 4.1
•
he feature net can be extended to other domains, includT
ing the recognition of three-dimensional objects. However,
the recognition of faces requires a different sort of model,
sensitive to configurations rather than to parts.
•
inally, we consider top-down influences on recognition.
F
The existence of these influences tells us that object recognition is not a self-contained process. Instead, knowledge external to object recognition is imported into and
clearly shapes the process.
D.F.’S LESIONS
A Lesions in subject D.F.
B Location of LOC in neurologically intact subjects
Panel A shows the location of the brain damage in D.F. Panel B shows the areas in the lateral occipital complex
(LOC) that are especially activated when neurologically healthy people are recognizing objects.
108 •
C H A P T E R F O U R Recognizing Objects
FIGURE 4.2
DRAWINGS FROM PATIENT D.F.
Line-drawing models
A
Drawn from the models
B
Drawn from memory
C
Patients who suffer from apperceptive agnosia can see, but they can’t organize the elements they see in order
to perceive an entire object. This deficit was evident when patient D.F. was asked to copy the drawings shown in
Panel A. Her attempts are shown in Panel B. The problem is not in her drawing ability, because D.F.’s performance
was much better (as shown in Panel C) when she was asked to draw the same forms from memory, rather than
from a model.
Dr. P. obviously can see, and he uses his (considerable) intelligence
to figure out what he is seeing. Nonetheless, his agnosia profoundly
disrupts his life. Sacks describes one incident in which Dr. P. failed to
put on his shoe, because he didn’t recognize it as a shoe. (In fact, Sacks
notes that at one point Dr. P. was confused about which object was his
shoe and which was his foot.) Then, at the end of their time together,
Sacks reports that Dr. P. “reached out his hand and took hold of his wife’s
head, tried to lift it off, to put it on. He had apparently mistaken his wife
for a hat!” (p. 11).
As these examples make clear, object recognition may not be a glamorous skill, but it is one that we all rely on for even our most ordinary
interactions with the world. What are the processes that make object
recognition possible?
Recognizing Objects
•
109
Recognition: Some Early Considerations
You’re obviously able to recognize a huge number of different patterns —
different objects (cats, cups, coats), various actions (crawling, climbing,
clapping), and different sorts of situations (crises, comedies). You can
also recognize many variations of each of these things. You recognize
cats standing up and cats sitting down, cats running and cats asleep.
And the same is true for recognition of most other patterns in your recog­
nition repertoire.
You also recognize objects even when the available information is
incomplete. For example, you can still recognize a cat if only its head
and one paw are visible behind a tree. You recognize a chair even when
someone is sitting on it, even though the person blocks much of the chair
from view.
All of this is true for print as well. You can recognize tens of thousands of
words, and you can recognize them whether the words are printed in large
type or small, italics or straight letters, UPPER CASE or lower. You can even
recognize handwritten words, for which the variation from one to the next
is huge.
These variations in the “stimulus input” provide our first indication
that object recognition involves some complexity. Another indication
comes from the fact that your recognition of various objects, print or other­
wise, is influenced by the context in which you encounter those objects.
Consider Figure 4.3. The middle character is the same in both words, but
the character looks more like an H in the word on the left and more like
an A in the word on the right. With this, you easily read the word on
the left as “THE” and not “TAE” and the word on the right as “CAT” and
not “CHT.”
Of course, object recognition is powerfully influenced by the stimulus
itself — that is, by the features that are in view. Processes directly shaped by
the stimulus are sometimes called “data driven” but are more commonly said
FIGURE 4.3
CONTEXT INFLUENCES PERCEPTION
You are likely to easily read this sequence as “THE CAT,” recognizing the middle
symbol as an H in one case and as an A in the other.
( after selfridge , 1955)
110 •
C H A P T E R F O U R Recognizing Objects
to involve bottom-up processing. The effect of context, however, reminds us
that recognition is also influenced by one’s knowledge and expectations. As a
result, your reading of Figure 4.3 is guided by your knowledge that “THE” and
“CAT” are common words but that “TAE” and “CHT” are not. This sort of
influence — relying on your knowledge — is sometimes called “concept-driven,”
and processes shaped by knowledge are said to involve top-down processing.
What mechanism underlies both the top-down and bottom-up influences?
In the next section, we’ll consider a classic proposal for what the mechanism
might be. We’ll then build on this base as we discuss more recent elaborations
of this proposal.
The Importance of Features
Common sense suggests that many objects can be recognized by virtue of their
parts. You recognize an elephant because you see the trunk, the thick legs, the
large body. You know a lollipop is a lollipop because you see the circle shape
on top of the straight stick. But how do you recognize the parts themselves?
How, for example, do you recognize the trunk on the elephant or the circle in
the lollipop? The answer may be simple: Perhaps you recognize the parts by
looking at their parts — such as the arcs that make up the circle in the lollipop,
or the (roughly) parallel lines that identify the elephant’s trunk.
To put this more generally, recognition might begin with the identification
of visual features in the input pattern — the vertical lines, curves, diagonals,
and so on. With these features appropriately catalogued, you can start assem­
bling the larger units. If you detect a horizontal together with a vertical, you
know you’re looking at a right angle; if you’ve detected four right angles,
you know you’re looking at a square.
This broad proposal lines up well with the neuroscience evidence we dis­
cussed in Chapter 3. There, we saw that specialized cells in the visual system
do seem to act as feature detectors, firing (producing an action potential)
whenever the relevant input (i.e., the appropriate feature) is in view. Also,
we’ve already noted that people can recognize many variations on the objects
they encounter — cats in different positions, A’s in different fonts or different
handwritings. An emphasis on features, though, might help with this point.
The various A’s, for example, differ from one another in overall shape, but
they do have certain things in common: two inwardly sloping lines and a
horizontal crossbar. Focusing on features, therefore, might allow us to con­
centrate on elements shared by the various A’s and so might allow us to
recognize A’s despite their apparent diversity.
The importance of features is also evident in data from visual search
tasks — tasks in which participants are asked to examine a display and to
judge whether a particular target is present in the display or not. This search
is remarkably efficient when someone is searching for a target defined by a
simple feature — for example, finding a vertical segment in a field of horizon­
tals or a green shape in a field of red shapes. But people are generally slower
THE VARIABILITY
OF STIMULI WE
RECOGNIZE
We recognize cats from the
side or the front, whether we
see them close up or far away.
Recognition: Some Early Considerations
•
111
FIGURE 4.4
VISUAL SEARCH
A
B
C
In Panel A, you can immediately spot the vertical, distinguished from the other shapes by just one feature. Likewise, in Panel B, you can immediately spot the lone green bar in the field of reds. But in Panel C, it takes longer
to find the one red vertical, because now you need to search for a combination of features — not just for red or
vertical, but for the one form that has both of these attributes.
TEST YOURSELF
1.What is the difference
between “bottomup” and “top-down”
processing?
2.What is the evidence
that features play a
special role in object
recognition?
in searching for a target defined as a combination of features (see Figure 4.4).
This is just what we would expect if feature analysis is an early step in your
analysis of the visual world — and separate from the step in which you com­
bine the features you’ve detected.
Further support for these claims comes from studies of brain damage.
At the start of the chapter, we mentioned apperceptive agnosia — a disorder
that involves an inability to assemble the various aspects of an input into an
organized whole. A related disorder, integrative agnosia, derives from damage
to the parietal lobe. Patients with this disorder appear relatively normal in
tasks requiring them simply to detect features in a display, but they are mark­
edly impaired in tasks that require them to judge how the features are bound
together to form complex objects. (See, for example, Behrmann, Peterson,
Moscovitch, & Suzuki, 2006; Humphreys & Riddoch, 2014; Robertson, Tre­
isman, Friedman-Hill, & Grabowecky, 1997. For related results, in which
transcranial magnetic stimulation was used to disrupt portions of the brain
in healthy individuals, see Ashbridge, Walsh, & Cowey, 1997.)
Word Recognition
Several lines of evidence, therefore, indicate that object recognition does
begin with the detection of simple features. Then, once this detection has
occurred, separate mechanisms are needed to put the features together,
assembling them into complete objects. But how does this assembly proceed,
so that we end up seeing not just the features but whole words — or Chihuahuas,
or fire hydrants? In tackling this question, it will be helpful to fill in some
more facts that we can then use as a guide to our theory building.
112 •
C H A P T E R F O U R Recognizing Objects
Factors Influencing Recognition
In many studies, participants have been shown stimuli for just a brief duration —
perhaps 20 or 30 ms (milliseconds). Older research did this by means of a
tachistoscope, a device designed to present stimuli for precisely controlled
amounts of time. More modern research uses computers, but the brief displays
are still called “tachistoscopic presentations.”
Each stimulus is followed by a post-stimulus mask — often, a random pat­
tern of lines and curves, or a random jumble of letters such as “XJDKEL.” The
mask interrupts any continued processing that participants might try to do
for the stimulus just presented. In this way, researchers can be certain that a
stimulus presented for (say) 20 ms is visible for exactly 20 ms and no longer.
Can people recognize these briefly visible stimuli? The answer depends on
many factors, including how familiar a stimulus is. If the stimulus is a word, we
can measure familiarity by counting how often that word appears in print, and
these counts are an excellent predictor of tachistoscopic recognition. In one early
experiment, Jacoby and Dallas (1981) showed participants words that were
either very frequent (appearing at least 50 times in every million printed words)
or infrequent (occurring only 1 to 5 times per million words of print). Participants
viewed these words for 35 ms, followed by a mask. Under these circumstances,
they recognized almost twice as many of the frequent words (see Figure 4.5A).
WORD FREQUENCY’S EFFECT ON WORD RECOGNITION
100
100
90
90
80
80
70
70
Percent Recognition
Percent Recognition
FIGURE 4.5
60
50
40
30
50
40
30
20
20
10
10
High-frequency
words
A
60
Low-frequency
words
B
Unprimed Primed
High-frequency
words
Unprimed Primed
Low-frequency
words
In one study, recognition was much more likely for words appearing often in print, in comparison to words appearing only rarely — an effect of frequency (Panel A). Similarly, words that had been viewed recently were more often
recognized, an effect of recency that in this case creates a benefit called “repetition priming” (Panel B).
( after jacoby & dallas , 1981)
Word Recognition
•
113
Another factor influencing recognition is recency of view. If partici­
pants view a word and then, a little later, view it again, they will recognize
the word more readily the second time around. The first exposure primes
the participant for the second exposure; more specifically, this is a case of
repetition priming.
As an example, participants in one study read a list of words aloud. The
participants were then shown a series of words in a tachistoscope. Some
of these words were from the earlier list and so had been primed; others
were unprimed. For words that were high in frequency, 68% of the unprimed
words were recognized, compared to 84% of the primed words. For words
low in frequency, 37% of the unprimed words were recognized, compared to
73% of the primed words (see Figure 4.5B; Jacoby & Dallas, 1981).
The Word-Superiority Effect
Figure 4.3 suggests that the recognition of a letter depends on its context —
and so an ambiguous letter is read as an A in one setting but an H in another
setting. But context also has another effect: Even when a letter is properly
printed and quite unambiguous, it’s easier to recognize if it appears within a
word than if it appears in isolation.
This result might seem paradoxical, because here we have a setting in
which it seems easier to do “more work” rather than “less” — and so you’re
more accurate in recognizing all the letters that make up a word (maybe
a total of five or six letters) than you are in recognizing just one letter on
its own. Paradoxical or not, this pattern is easy to demonstrate, and the
advantage for perceiving letters-in-context is called the word-superiority
effect (WSE).
The WSE is demonstrated with a “two-alternative, forced-choice” pro­
cedure. For example, in some trials we might present a single letter — let’s
say K — followed by a post-stimulus mask, and follow that with a question:
“Which of these was in the display: an E or a K?” In other trials, we might
present a word — let’s say “DARK” — followed by a mask, followed by a
question: “Which of these was in the display: an E or a K?”
Note that participants have a 50-50 chance of guessing correctly in either
of these situations, and so any contribution from guessing is the same for the
letters as it is for the words. Also, for the word stimulus, both of the letters
we’ve asked about are plausible endings for the stimulus; either ending would
create a common word (“DARE” or “DARK”). Therefore, participants who
saw only part of the display (perhaps “DAR”) couldn’t use their knowledge
of the language to figure out the display’s final letter. In order to choose
between E and K, therefore, participants really need to have seen the relevant
letter — and that is exactly what we want.
In this procedure, accuracy rates are reliably higher in the word condi­
tion. Apparently, recognizing an entire word is easier than recognizing iso­
lated letters (see Figure 4.6; Johnston & McClelland, 1973; Reicher, 1969;
Rumelhart & Siple, 1974; Wheeler, 1970).
114 •
C H A P T E R F O U R Recognizing Objects
Percent correct
80
70
FIGURE 4.6
60
50
Single letters Entire words
Stimulus type
THE WORD-SUPERIORITY EFFECT
The word-superiority effect is usually demonstrated with a twoalternative forced-choice procedure (which means that a participant
can get a score of 50% just by guessing randomly). Performance is
much better if the target letter is shown in context — within an entire
word — than if it is shown on its own.
( after johnston & m c clelland , 1973)
Degree of Well-Formedness
As it turns out, though, the term “word-superiority effect” may be mis­
leading, because we don’t need words to produce the pattern evident in
Figure 4.6. We get a similar effect if we present participants with letter
strings like “FIKE” or “LAFE.” These letter strings are not English words
and they’re not familiar, but they look like English strings and (related)
are easy to pronounce. And, crucially, strings like these produce a context
effect, with the result that letters in these contexts are easier to identify than
letters alone.
This effect occurs, though, only if the context is of the right sort. There’s
no context benefit if we present a string like “HZYE” or “SBNE.” An E pre­
sented within these strings will not show the word-superiority effect — that
is, it won’t be recognized more readily than an E presented in isolation.
A parallel set of findings emerge if, instead of asking participants to detect
specific letters, we ask them to report all of what they have seen. A letter
string like “HZYE” is extremely hard to recognize if presented briefly. With
a stimulus like this and, say, a 30-ms exposure, participants may report that
they only saw a flash and no letters at all; at best, they may report a letter or
two. But with the same 30-ms exposure, participants will generally recognize
(and be able to report) strings like “FIKE” or “LAFE,” although they do even
better if the stimuli presented are actual, familiar words.
How should we think about these findings? One approach emphasizes the
statistically defined regularities in English spelling. Specifically, we can work
through a dictionary, counting how often (for example) the letter combination
“FI” occurs, or the combination “LA,” or “HZ.” We can do the same for threeletter sequences (“FIK,” “LAF,” and so on). These counts will give us a tally that
reveals which letter combinations are more probable in English spelling and
Word Recognition
•
115
which are less probable. We can then use this tally to evaluate new strings —
asking, for any string, whether its letter sequences are high-probability ones
(occurring often) or low-probability (occurring rarely).
These statistical measures allow us to evaluate how “well formed” a letter
string is — that is, how well the letter sequence conforms to the usual spelling
patterns of English — and well-formedness is a good predictor of word rec­
ognition: The more English-like the string is, the easier it will be to recognize
that string, and also the greater the context benefit the string will produce.
This well-documented pattern has been known for more than a century (see,
e.g., Cattell, 1885) and has been replicated in many studies (Gibson, Bishop,
Schiff, & Smith, 1964; Miller, Bruner, & Postman, 1954).
Making Errors
TEST YOURSELF
3.What is repetition
priming, and how
is it demonstrated?
4. What procedure demonstrates the wordsuperiority effect?
5. What’s the evidence
that word perception
is somehow governed
by the rules of ordinary
spelling?
Let’s recap some important points. First, it seems that a letter will be easier
to recognize if it appears in a well-formed sequence, but not if it appears in
a random sequence. Second, well-formed strings are, overall, easier to per­
ceive than ill-formed strings; this advantage remains even if the well-formed
strings are made-up ones that you’ve never seen before (strings like “HAKE”
or “COTER”). All of these facts suggest that you somehow are using your
knowledge of spelling patterns when you look at, and recognize, the words
you encounter — and so you have an easier time with letter strings that con­
form to these patterns, compared to strings that do not.
The influence of spelling patterns is also evident in the mistakes you make.
With brief exposures, word recognition is good but not perfect, and the errors
that occur are systematic: There’s a strong tendency to misread less-common
letter sequences as if they were more-common patterns. So, for example,
“TPUM” is likely to be misread as “TRUM” or even “DRUM.” But the reverse
errors are rare: “DRUM” is unlikely to be misread as “TRUM” or “TPUM.”
These errors can sometimes be quite large — so that someone shown
“TPUM” might instead perceive “TRUMPET.” But, large or small, the errors
show the pattern described: Misspelled words, partial words, or nonwords
are read in a way that brings them into line with normal spelling. In effect,
people perceive the input as being more regular than it actually is. Once
again, therefore, our recognition seems to be guided by (or, in this case, mis­
guided by) some knowledge of spelling patterns.
Feature Nets and Word Recognition
What lies behind this broad pattern of evidence? What are the processes in­
side of us that lead to the findings we’ve described? Psychologists’ under­
standing of these points grows out of a theory published many years ago
(Selfridge, 1959). Let’s start with that theory, and then use it as our base
as we look at more modern work. (For a glimpse of some of the modern
research, including work that links theorizing to neuroscience, see Carreiras,
Armstrong, Perea, & Frost, 2014.)
116 •
C H A P T E R F O U R Recognizing Objects
COGNITION
outside the lab
Font
You encounter printed material in a variety of
the students read these facts in a clear font (Arial
formats and in a wide range of fonts. You also
printed in pure black), and half read the facts in less
come across hand-written material, and of course
clear font (e.g., Bodoni MT, printed in 60% grayscale).
people differ enormously in their handwriting.
When tested later, participants who’d seen the
Despite all this variety, you’re able to read almost
fluent print remembered 73% of the facts; par-
everything you see — somehow rising above the
ticipants who’d seen the less fluent print recalled
variations from one bit to the next.
86% of the facts. What was going on here? We’ll
These variations do matter, however. Some
see in Chapter 6 that memory is promoted by
people’s handwriting is an almost impenetrable
active engagement with the to-be-remembered
scrawl; some fonts are difficult to decipher. Even
materials, and it seems that the somewhat
if we step away from these extremes and only
obscure font promoted that sort of engagement —
consider cases in which you can figure out what’s
and so created what (in Chapter 6) we’ll refer to
on the page, poor handwriting or an obscure
as “desirable difficulty” in the learning process.
font can make your reading less fluent. How this
What about other aspects of formatting?
drop in fluency matters, though, depends on the
We’ve discussed the individual features that you
circumstances.
use in recognizing letters, but it turns out that
In one study, college students read a pas-
you’re also sensitive to a word’s overall shape. This
sage printed either in a clear font (Times New
is one of the reasons WHY IT IS MORE DIFFICULT
Roman) or in a difficult font (italicized Juice ITC).
TO READ CAPITALIZED TEXT. Capitalized words
Students in both groups were then asked to rate
all have the same rectangular shape; gone are the
the intelligence of the author who’d written the
portions of the letter that hang belong the line —
passage (Oppenheimer, 2006). Students who
the so-called descenders, like the bottom tail on a
read the less clear font rated the author as less
g or a j. Also gone are the portions of the letters
intelligent; apparently, they had noticed that
that stick up (ascenders), like the top of an h or
their reading wasn’t fluent but didn’t realize the
an l, or the dot over an i. Your reading slows down
problem was in the font. Instead, they decided
when these features aren’t available, so it’s slower
that the lack of fluency was the author’s fault:
IF YOU READ BLOCK CAPITALS compared to the
They decided that the author hadn’t been clear
normal pattern of print.
enough in composing the passage, and there-
Are there practical lessons here? In some
cases, you might prefer the look of block capitals;
fore was less intelligent!
an
but if so, be aware that this format slows reading a
advantage for a (slightly) obscure font (Diemond-
bit. In choosing a font, you should probably avoid
Yauman, Oppenheimer, & Vaughan, 2011). Col-
the obscure styles (unless you want less-fluent
lege students were asked to read made-up facts
reading!), but notice that a moderately challeng-
about space aliens — for example, that the Nor-
ing font can actually help readers to process and
gletti are 2 ft tall and eat flower petals. Half of
remember what you’ve written.
Another
experiment,
though,
showed
Feature Nets and Word Recognition
•
117
The Design of a Feature Net
Imagine that we want to design a system that will recognize the word
“CLOCK” whenever it is in view. How might our “CLOCK” detector work?
One option is to “wire” this detector to a C-detector, an L-detector, an
O-detector, and so on. Then, whenever these letter detectors are activated,
this would activate the word detector. But what activates the letter detectors?
Maybe the L-detector is “wired” to a horizontal-line detector and also a
vertical-line detector, as shown in Figure 4.7. When these feature detectors
are activated, this activates the letter detector.
The idea is that there could be a network of detectors, organized in layers.
The “bottom” layer is concerned with features, and that is why networks of
this sort are often called feature nets. As we move “upward” in the network,
each subsequent layer is concerned with larger-scale objects; using the term
we introduced earlier, the flow of information would be bottom-up — from
the lower levels toward the upper levels.
But what does it mean to “activate” a detector? At any point in time, each
detector in the network has a particular activation level, which reflects the
status of the detector at that moment — roughly, how energized the detec­
tor is. When a detector receives some input, its activation level increases. A
strong input will increase the activation level by a lot, and so will a series of
weaker inputs. In either case, the activation level will eventually reach the
detector’s response threshold, and at that point the detector will fire — that is,
send its signal to the other detectors to which it is connected.
FIGURE 4.7
A SIMPLE FEATURE NET
Word
detector
CLOCK
C
L
O
C
K
Letter
detectors
Feature
detectors
An example of a feature net. Here, the feature detectors respond to simple elements in the visual input. When the appropriate feature detectors are activated,
they trigger a response in the letter detectors. When these are activated, in
turn, they can trigger a response in a higher-level detector, such as a detector
for an entire word.
118 •
C H A P T E R F O U R Recognizing Objects
These points parallel our description of neurons in Chapter 2, and that’s
no accident. If the feature net is to be a serious candidate for how humans
recognize patterns, then it has to use the same sorts of building blocks that
the brain does. However, let’s be careful not to overstate this point: No one
is suggesting that detectors are neurons or even large groups of neurons.
Instead, detectors probably involve complex assemblies of neural tissue.
Nonetheless, it’s plainly attractive that the hypothesized detectors in the fea­
ture net function in a way that’s biologically sensible.
Within the net, some detectors will be easier to activate than others —
that is, some will require a strong input to make them fire, while oth­
ers will fire even with a weak input. This difference is created in part by
how activated each detector is to begin with. If the detector is moderately
activated at the start, then only a little input is needed to raise the activa­
tion level to threshold, and so it will be easy to make this detector fire. If
a detector is not at all activated at the start, then a strong input is needed
to bring the detector to threshold, and so it will be more difficult to make
this detector fire.
What determines a detector’s starting activation level? As one factor,
detectors that have fired recently will have a higher activation level (think
of it as a “warm-up” effect). In addition, detectors that have fired frequently
in the past will have a higher activation level (think of it as an “exercise”
effect). Overall, then, activation level is dependent on principles of recency
and frequency.
We now can put these mechanisms to work. Why are frequent words in
the language easier to recognize than rare words? Frequent words, by defini­
tion, appear often in the things you read. Therefore, the detectors needed for
recognizing these words have been frequently used, so they have relatively
high levels of activation. Thus, even a weak signal (e.g., a brief or dim presen­
tation of the word) will bring these detectors to their response threshold and
will be enough to make them fire. As a result, the word will be recognized
even with a degraded input.
Repetition priming is explained in similar terms. Presenting a word once
will cause the relevant detectors to fire. Once they’ve fired, activation levels
will be temporarily lifted (because of recency of use). Therefore, only a weak
signal will be needed to make the detectors fire again. As a result, the word
will be more easily recognized the second time around.
The Feature Net and Well-Formedness
The net we’ve described so far cannot, however, explain all of the data.
Consider the effects of well-formedness — for instance, the fact that people
are able to read letter strings like “PIRT” or “HICE” even when those strings
are presented very briefly (or dimly or in low contrast), but not strings like
“ITPR” or “HCEI.” How can we explain this finding? One option is to add
another layer to the net, a layer filled with detectors for letter combinations.
Feature Nets and Word Recognition
•
119
FIGURE 4.8
BIGRAM DETECTORS
Word
detector
Bigram
detectors
Letter
detectors
Feature
detectors
It seems plausible that the network includes a layer of bigram detectors
between the letter detectors and word detectors.
Thus, in Figure 4.8, we’ve added a layer of bigram detectors — detectors of
letter pairs. These detectors, like all the rest, will be triggered by lower-level
detectors and send their output to higher-level detectors. And just like any
other detector, each bigram detector will start out with a certain activation
level, influenced by the frequency with which the detector has fired in the past
and by the recency with which it has fired.
This turns out to be all the theory we need. You have never seen the
sequence “HICE” before, but you have seen the letter pair HI (in “HIT,”
“HIGH,” or “HILL”) and the pair CE (“FACE,” “MICE,” “JUICE”).
The detectors for these letter pairs, therefore, have high activation levels at
the start, so they don’t need much additional input to reach their thresh­
old. As a result, these detectors will fire with only weak input. That will
make the corresponding letter combinations easy to recognize, facilitating
the recognition of strings like “HICE.” None of this is true for “IJPV”
or “RSFK.” Because none of these letter combinations are familiar, these
strings will receive no benefits from priming. As a result, a strong input will
be needed to bring the relevant detectors to threshold, and so these strings
will be recognized only with difficulty. (For more on bigram detectors
and how they work, see Grainger, Rey, & Dufau, 2008; Grainger &
Whitney, 2004; Whitney, 2001. For some complications, see Rayner &
Pollatsek, 2011.)
120 •
C H A P T E R F O U R Recognizing Objects
Recovery from Confusion
Imagine that we present the word “CORN” for just 20 ms. In this setting, the
visual system has only a limited opportunity to analyze the input, so it’s possible that
you’ll miss some of the input’s features. For example, let’s imagine that the second
letter in this word — the O — is hard to see, so that only the bottom curve is detected.
This partial information invites confusion. If all you know is “the second
letter had a bottom curve,” then perhaps this letter was an O, or perhaps it
was a U, or a Q, or maybe an S. Figure 4.9 shows how this would play out
in terms of the network. We’ve already said that you detected the bottom
curve, and that means the “bottom-curve detector” is activated. This detec­
tor, in turn, provides input to the O-detector and also to the detectors for U,
FIGURE 4.9
THE VISUAL PROCESSING PATHWAYS
Word
detectors
Bigram
detectors
Letter
detectors
Feature
detectors
Stimulus
input
If “CORN” is presented briefly, not all of its features will be detected. Imagine
that only the bottom curve of the O is detected, not the O’s top or sides. This will
(weakly) activate the O-detector, but it will also activate the detectors of various
other letters having a bottom curve, including U, Q, and S. This will, in turn, send
weak activation to the appropriate bigram detectors. The CO-detector, however,
is well primed and so is likely to respond even though it is receiving only a weak
input. The other bigram detectors (for CQ or CS) are less well primed and so will
not respond to this weak input. Therefore, “CORN” will be correctly perceived,
despite the confusion at the letter level caused by the weak signal.
Feature Nets and Word Recognition
•
121
Q, and S, and so activation in this feature detector causes activation in all of
these letter detectors.
Of course, each of these letter detectors is wired so that it can also
receive input from other feature detectors. (And so usually the O-detector also
gets input from detectors for left curves, right curves, and top curves.) We’ve
already said, though, that with this brief input these other features weren’t
detected this time around. As a result, the O-detector will only be weakly
activated (because it’s not getting its usual full input), and the same is true for
the detectors for U, Q, and S.
In this situation, therefore, the network has partial information at the
feature level (because only one of the O’s features was detected), and this
leads to confusion at the letter level: Too many letter detectors are firing
(because the now-activated bottom-curve detector is wired to all of them).
And, roughly speaking, all of these letter detectors are firing in a fashion that
signals uncertainty, because they’re each receiving input from only one of
their usual feature detectors.
The confusion continues in the information sent upward from the let­
ter level to the bigram level. The detector for the CO bigram will receive a
strong signal from the C-detector (because the C was clearly visible) but only
a weak signal from the O-detector (because the O wasn’t clearly visible).
The CU-detector will get roughly the same input — a strong signal from the
C-detector and a weak signal from the U-detector. Likewise for the CQ- and
CS-detectors. In other words, we can imagine that the signal being sent from
the letter detectors is “maybe CO or maybe CU or maybe CQ or maybe CS.”
The confusion is, however, sorted out at the bigram level. All four bigram
detectors in this situation are receiving the same input — a strong signal from
one of their letters and a weak signal from the other. But the four detectors don’t
all respond in the same way. The CO-detector is well primed (because this is
a frequent pattern), so the activation it’s receiving will probably be enough to
fire this (primed) detector. The CU-detector is less primed (because this is a less
frequent pattern); the CQ- and CS-detectors, if they even exist, are not primed
at all. The input to these latter detectors is therefore unlikely to activate them —
because, again, they’re less well primed and so won’t respond to this weak input.
What will be the result of all this? The network was “under-stimulated”
at the feature level (with only a subset of the input’s features detected) and
therefore confused at the letter level (with too many detectors firing). But
then, at the bigram level, it’s only the CO-detector that fires, because at this
level it is the detector (because of priming) most likely to respond to the weak
input. Thus, in a totally automatic fashion, the network recovers from its
own confusion and, in this case, avoids an error.
Ambiguous Inputs
Look again at Figure 4.3. The second character is exactly the same as the fifth,
but the left-hand string is perceived as “THE” (and the character is identified as
an H) and the right-hand string is perceived as “CAT” (and the character as an A).
122 •
C H A P T E R F O U R Recognizing Objects
What’s going on here? In the string on the left, the initial T is clearly in
view, and so presumably the T-detector will fire strongly in response. The
next character in the display will probably trigger some of the features nor­
mally associated with an A and some normally associated with an H. This
will cause the A-detector to fire, but only weakly (because only some of the
A’s features are present), and likewise for the H-detector. At the letter level,
then, there will be uncertainty about what this character is.
What happens next, though, follows a by-now familiar logic: With only
weak activation of the A- and H-detectors, only a moderate signal will be
sent upward to the TH- and TA-detectors. Likewise, it seems plausible that
only a moderate signal will be sent to the THE- and TAE-detectors at the
word level. But, of course, the THE-detector is enormously well primed;
if there is a TAE-detector, it would be barely primed, since this is a string
that’s rarely encountered. Thus, the THE- and TAE-detectors might be
receiving similar input, but this input is sufficient only for the (well-primed)
THE-detector, so only it will respond. In this way, the net will recognize
the ambiguous pattern as “THE,” not “TAE.” (The same logic applies,
of course, to the ambiguous pattern on the right, perceived as “CAT,”
not “CHT.”)
A similar explanation will handle the word-superiority effect (see, e.g.,
Rumelhart & Siple, 1974). To take a simple case, imagine that we present
the letter A in the context “AT.” If the presentation is brief enough, par­
ticipants may see very little of the A, perhaps just the horizontal crossbar.
This wouldn’t be enough to distinguish among A, F, or H, and so all these
letter detectors would fire weakly. If this were all the information the par­
ticipants had, they’d be stuck. But let’s imagine that the participants did
perceive the second letter in the display, the T. It seems likely that the AT
bigram is much better primed than the FT or HT bigrams. (That’s because
you often encounter words like “CAT” or “BOAT”; words like “SOFT”
or “HEFT” are used less frequently.) Therefore, the weak firing of the
A-detector would be enough to fire the AT bigram detector, while the weak
firing for the F and H might not trigger their bigram detectors. In this way,
a “choice” would be made at the bigram level that the input was “AT” and
not something else. Once this bigram has been detected, answering the
question “Was there an A or an F in the display?” is easy. In this way, the
letter will be better detected in context than in isolation. This isn’t because
context enables you to see more; instead, context allows you to make better
use of what you see.
Recognition Errors
There is, however, a downside to all this. Imagine that we present the string
“CQRN” to participants. If the presentation is brief enough, the partici­
pants will register only a subset of the string’s features. Let’s imagine that
they register only the bottom bit of the string’s second letter. This detec­
tion of the bottom curve will weakly activate the Q-detector and also the
Feature Nets and Word Recognition
•
123
FIGURE 4.10
RECOGNITION ERRORS
Word
detectors
Bigram
detectors
Letter
detectors
Feature
detectors
Stimulus
input
If “CQRN” is presented briefly, not all of its features will be detected. Perhaps
only the bottom curve of the Q is detected, and this will weakly activate various other letters having a bottom curve, including O, U, and S. However, the
same situation would result from a brief presentation of “CORN” (as shown in
Figure 4.9); therefore, by the logic we have already discussed, this stimulus is
likely to be misperceived as “CORN.”
U-detector and the O-detector. The resulting pattern of network activation
is shown in Figure 4.10.
Of course, the pattern of activation here is exactly the same as it was in
Figure 4.9. In both cases, perceivers have seen the features for the C, R, and
N and have only seen the second letter’s bottom curve. And we’ve already
walked through the network’s response to this feature pattern: This con­
figuration will lead to confusion at the letter level, but the confusion will
get sorted out at the bigram level, with the (primed) CO-detector respond­
ing to this input and other (less well primed) detectors not responding.
As a result, the stimulus will be (mis)identified as “CORN.” In the situa­
tion described in Figure 4.9, the stimulus actually was “CORN,” and so
the dynamic built into the net aids performance, allowing the network to
recover from its initial confusion. In the case we’re considering now (with
124 •
C H A P T E R F O U R Recognizing Objects
“CQRN” as the stimulus), the exact same dynamic causes the network to
misread the stimulus.
This example helps us understand how recognition errors occur and why
those errors tend to make the input look more regular than it really is. The
basic idea is that the network is biased, favoring frequent letter combinations
over infrequent ones. In effect, the network operates on the basis of “when
in doubt, assume that the input falls into the frequent pattern.” The reason,
of course, is simply that the detectors for the frequent pattern are well primed —
and therefore easier to trigger.
Let’s emphasize, though, that the bias built into the network facilitates perception if the input is, in fact, a frequent word, and these (by
definition) are the words you encounter most of the time. The bias will
pull the network toward errors if the input happens to have an unusual
spelling pattern, but (by definition) these inputs are less common in your
experience. Hence, the network’s bias helps perception more often than
it hurts.
Distributed Knowledge
We’ve now seen many indications that the network’s functioning is guided by
knowledge of spelling patterns. This is evident in the fact that letter strings
are easier to recognize if they conform to normal spelling. The same point
is evident in the fact that letter strings provide a context benefit (the WSE)
only if they conform to normal spelling. Even more evidence comes from the
fact that when errors occur, they “shift” the perception toward patterns of
normal spelling.
To explain these results, we’ve suggested that the network “knows” (for
example) that CO is a common bigram in English, while CF is not, and
also “knows” that THE is a common sequence but TAE is not. The net­
work seems to rely on this “knowledge” in “choosing” its “interpretation” of
unclear or ambiguous inputs. Similarly, the network seems to “expect”
certain patterns and not others, and is more efficient when the input lines up
with those “expectations.”
Obviously, we’ve wrapped quotations around several of these words to
emphasize that the sense in which the net “knows” facts about spelling,
or the sense in which it “expects” things or makes “interpretations,” is a
little peculiar. In reality, knowledge about spelling patterns isn’t explicitly
stored anywhere in the network. Nowhere within the net is there a sentence
like “CO is a common bigram in English; CF is not.” Instead, this memory
(if we even want to call it that) is manifest only in the fact that the COdetector happens to be more primed than the CF-detector. The CO-detector
doesn’t “know” anything about this advantage, nor does the CF-detector
know anything about its disadvantage. Each one simply does its job, and
in the course of doing their jobs, sometimes a “competition” will take
place between these detectors. (This sort of competition was illustrated in
Feature Nets and Word Recognition
•
125
Figures 4.9 and 4.10.) When these competitions occur, they’ll be “decided” by
activation levels: The better-primed detector will be more likely to respond
and therefore will be more likely to influence subsequent events. That’s the
entire mechanism through which these “knowledge effects” arise. That’s
how “expectations” or “inferences” emerge — as a direct consequence of
the activation levels.
To put this into technical terms, the network’s “knowledge” is not locally
represented anywhere; it isn’t stored in a particular location or built into
a specific process. As a result, we cannot look just at the level of priming
in the CO-detector and conclude that this detector represents a frequently
seen bigram. Nor can we look at the CF-detector and conclude that it
represents a rarely seen bigram. Instead, we need to look at the relationship between these priming levels, and we also need to look at how this
relationship will lead to one detector being more influential than the
other. In this way, the knowledge about bigram frequencies is contained
within the network via a distributed representation; it’s knowledge, in
other words, that’s represented by a pattern of activations that’s distrib­
uted across the network and detectable only if we consider how the entire
network functions.
What may be most remarkable about the feature net, then, lies in how
much can be accomplished with a distributed representation, and thus with
simple, mechanical elements correctly connected to one another. The net
appears to make inferences and to know the rules of English spelling. But
the actual mechanics of the net involve neither inferences nor knowledge (at
least, not in any conventional sense). You and I can see how the inferences
unfold by taking a bird’s-eye view and considering how all the detectors work
together as a system. But nothing in the net’s functioning depends on the
bird’s-eye view. Instead, the activity of each detector is locally determined —
influenced by just those detectors feeding into it. When all these detectors
work together, though, the result is a process that acts as if it knows the rules.
But the rules themselves play no role in guiding the network’s moment-bymoment activities.
Efficiency versus Accuracy
One other point about the network needs emphasis: The network does
make mistakes, misreading some inputs and misinterpreting some pat­
terns. As we’ve seen, though, these errors are produced by exactly
the same mechanisms that are responsible for the network’s main
advantages — its ability to deal with ambiguous inputs, for example, or to
recover from confusion. Perhaps, therefore, we should view the errors as
the price you pay in order to gain the benefits associated with the net:
If you want a mechanism that’s able to deal with unclear or partial inputs,
you have to live with the fact that sometimes the mechanism will make
errors.
126 •
C H A P T E R F O U R Recognizing Objects
But do you really need to pay this price? After all, outside of the lab
you’re unlikely to encounter fast-paced tachistoscopic inputs. Instead, you
see stimuli that are out in view for long periods of time, stimuli that you can
inspect at your leisure. Why, therefore, don’t you take the moment to scruti­
nize these inputs so that you can rely on fewer inferences and assumptions,
and in that way gain a higher level of accuracy in recognizing the objects
you encounter?
The answer is straightforward. To maximize accuracy, you could, in
principle, scrutinize every character on the page. That way, if a character
were missing or misprinted, you would be sure to detect it. But the cost
associated with this strategy would be intolerable. Reading would be un­
speakably slow (partly because the speed with which you move your eyes
is relatively slow — no more than four or five eye movements per second).
In contrast, it’s possible to make inferences about a page with remark­
able speed, and this leads readers to adopt the obvious strategy: They read
some of the letters and make inferences about the rest. And for the most
part, those inferences are safe — thanks to the simple fact that our language
(like most aspects of our world) contains some redundncies, so that one
doesn’t need every lettr to identify what a wrd is; oftn the missng letter is
perfctly predctable from the contxt, virtually guaranteeing that inferences
will be correct.
TEST YOURSELF
6. H
ow does a feature
net explain the wordfrequency effect?
7. How does a feature
net explain the types
of errors people make
in recognizing words?
8. What are the benefits,
and what are the
costs, associated
with the feature net’s
functioning?
Descendants of the Feature Net
We mentioned early on that we were discussing the “classic” version of the
feature net. This discussion has enabled us to bring a number of themes into
view — including the trade-off between efficiency and accuracy and the idea
of distributed knowledge built into a network’s functioning.
Over the years, though, researchers have offered improvements on
this basic conceptualization, and in the next sections we’ll consider three
of their proposals. All three preserve the idea of a network of intercon­
nected detectors, but all three extend this idea in important ways. We’ll
look first at a proposal that highlights the role of inhibitory connections
among detectors. Then we’ll turn to a proposal that applies the network
idea to the recognition of complex three-dimensional objects. Finally, we’ll
consider a proposal that rests on the idea that your ability to recognize
objects may depend on your viewing perspective when you encounter
those objects.
The McClelland and Rumelhart Model
In the network proposal we’ve considered so far, activation of one detec­
tor serves to activate other detectors. Other models involve a mechanism
through which detectors can inhibit one another, so that the activation of one
detector can decrease the activation in other detectors.
Descendants of the Feature Net
•
127
FIGURE 4.11
N ALTERNATIVE CONCEPTION OF THE
A
FEATURE NETWORK
The McClelland and Rumelhart (1981) pattern-recognition model includes
both excitatory connections (indicated by red arrows) and inhibitory connections (indicated by connections with dots). Connections within a specific level
are also possible — so that, for example, activation of the “TRIP” detector will
inhibit the detectors for “TRAP,” “TAKE,” and “TIME.”
One highly influential model of this sort was proposed by McClelland and
Rumelhart (1981); a portion of their model is illustrated in Figure 4.11. This net­
work, like the one we’ve been discussing, is better able to identify well-formed
strings than irregular strings; this net is also more efficient in identifying charac­
ters in context as opposed to characters in isolation. However, several attributes
of this net make it possible to accomplish all this without bigram detectors.
In Figure 4.11, excitatory connections — connections that allow one detec­
tor to activate its neighbors — are shown as red arrows; for example, detec­
tion of a T serves to “excite” the “TRIP” detector. Other connections are
inhibitory, and so (for example) detection of a G deactivates, or inhibits,
the “TRIP” detector. These inhibitory connections are shown in the figure
with dots. In addition, this model allows for more complicated signaling
than we’ve used so far. In our discussion, we have assumed that lower-level
detectors trigger upper-level detectors, but not the reverse. The flow of infor­
mation, it seemed, was a one-way street. In the McClelland and Rumelhart
128 •
C H A P T E R F O U R Recognizing Objects
model, though, higher-level detectors (word detectors) can influence lowerlevel detectors, and detectors at any level can also influence other detectors
at the same level (e.g., letter detectors can inhibit other letter detectors; word
detectors can inhibit other word detectors).
To see how this would work, let’s say that the word “TRIP” is briefly shown,
allowing a viewer to see enough features to identify only the R, I, and P.
Detectors for these letters will therefore fire, in turn activating the detector for
“TRIP.” Activation of this word detector will inhibit the firing of other word
detectors (e.g., detectors for “TRAP” and “TAKE”), so that these other words
are less likely to arise as distractions or competitors with the target word.
At the same time, activation of the “TRIP” detector will also excite
the detectors for its component letters — that is, detectors for T, R, I,
and P. The R-, I-, and P-detectors, we’ve assumed, were already firing, so this
extra activation “from above” has little impact. But the T-detector wasn’t
firing before. The relevant features were on the scene but in a degraded form
(thanks to the brief presentation), and this weak input was insufficient to
trigger an unprimed detector. But once the excitation from the “TRIP” detec­
tor primes the T-detector, it’s more likely to fire, even with a weak input.
In effect, then, activation of the word detector for “TRIP” implies that
this is a context in which a T is quite likely. The network therefore responds
to this suggestion by “preparing itself” for a T. Once the network is suitably
prepared (by the appropriate priming), detection of this letter is facilitated. In
this way, the detection of a letter sequence (the word “TRIP”) makes the net­
work more sensitive to elements that are likely to occur within that sequence.
That is exactly what we need in order for the network to be responsive to the
regularities of spelling patterns.
Let’s also note that the two-way communication that’s in play here fits
well with how the nervous system operates: Neurons in the eyeballs send
activation to the brain but also receive activation from the brain; neurons
in the lateral geniculate nucleus (LGN) send activation to the visual cortex
but also receive activation from the cortex. Facts like these make it clear that
visual processing is not a one-way process, with information flowing simply
from the eyes toward the brain. Instead, signaling occurs in both an ascend­
ing (toward the brain) and a descending (away from the brain) direction, just
as the McClelland and Rumelhart model claims.
Recognition by Components
The McClelland and Rumelhart model — like the feature net we started
with — was designed initially as an account of how people recognize printed
language. But, of course, we recognize many objects other than print, includ­
ing the three-dimensional objects that fill our world — chairs and lamps and
cars and trees. Can these objects also be recognized by a feature network?
The answer turns out to be yes.
Consider a network theory known as the recognition by components
(RBC) model (Hummel & Biederman, 1992; Hummel, 2013). This model
Descendants of the Feature Net
•
129
FIGURE 4.12
GEONS
Geons
Objects
2
4
2
1
3
3
3
5
5
3
4
3
2
5
5
5
3
A
1
3
B
Panel A shows five different geons; Panel B shows how these geons can
be assembled into objects. The numbers in Panel B identify the specific
geons — for example, a bucket contains Geon 5 top-connected to Geon 3.
includes several important innovations, one of which is the inclusion of an
intermediate level of detectors, sensitive to geons (short for “geometric
ions”). The idea is that geons might serve as the basic building blocks of all
the objects we recognize — geons are, in essence, the alphabet from which
all objects are constructed.
Geons are simple shapes, such as cylinders, cones, and blocks (see
Figure 4.12A), and according to Biederman (1987, 1990), we only need 30
or so different geons to describe every object in the world, just as 26 letters
are all we need to spell all the words of English. These geons can be com­
bined in various ways — in a top-of relation, or a side-connected relation,
and so on — to create all the objects we perceive (see Figure 4.12B).
The RBC model, like the other networks we’ve been discussing, uses
a hierarchy of detectors. The lowest-level detectors are feature detectors,
which respond to edges, curves, angles, and so on. These detectors in turn
activate the geon detectors. Higher levels of detectors are then sensitive
to combinations of geons. More precisely, geons are assembled into com­
plex arrangements called “geon assemblies,” which explicitly represent
the relations between geons (e.g., top-of or side-connected). These assemblies,
130 •
C H A P T E R F O U R Recognizing Objects
finally, activate the object model, a representation of the complete, recog­
nized object.
The presence of the geon and geon-assembly levels within this hierarchy of­
fers several advantages. For one, geons can be identified from virtually any angle
of view. As a result, recognition based on geons is viewpoint-independent. Thus,
no matter what your position is relative to a cat, you’ll be able to identify its
geons and identify the cat. Moreover, it seems that most objects can be recog­
nized from just a few geons. As a consequence, geon-based models like RBC
can recognize an object even if many of the object’s geons are hidden from view.
Recognition via Multiple Views
A number of researchers have offered a different approach to object recogni­
tion (Hayward & Williams, 2000; Tarr, 1995; Tarr & Bülthoff, 1998; Vuong
& Tarr, 2004; Wallis & Bülthoff, 1999). They propose that people have stored
in memory a number of different views of each object they can recognize: an
image of what a cat looks like when viewed head-on, an image of what it looks
like from the left, and so on. According to this perspective, you’ll recognize
Felix as a cat only if you can match your current view of Felix with one of these
remembered views. But the number of views in memory is limited — maybe a
half dozen or so — and so, in many cases, your current view won’t line up with
any of the available images. In that situation, you’ll need to “rotate” the cur­
rent view to bring it into alignment with one of the views in memory, and this
mental rotation will cause a slight delay in the recognition.
The key, then, is that recognition sometimes requires mental rotation, and
as a result it will be slower from some viewpoints than from others. In other
words, the speed of recognition will be viewpoint-dependent, and a growing
body of data confirms this claim. We’ve already noted that you can recognize
objects from many different angles, and your recognition is generally fast.
However, data indicate that recognition is faster from some angles than oth­
ers, in a way that’s consistent with this multiple-views proposal.
According to this perspective, how exactly does viewpoint-dependent rec­
ognition proceed? One proposal resembles the network models we’ve been
discussing (Riesenhuber & Poggio, 1999, 2002; Tarr, 1999). In this proposal,
there is a hierarchy of detectors, with each successive layer within the network
concerned with more complex aspects of the whole. Thus, low-level detectors
respond to lines at certain orientations; higher-level detectors respond to cor­
ners and notches. At the top of the hierarchy are detectors that respond to
the sight of whole objects. It is important, though, that these detectors each
represent what the object looks like from a particular vantage point, and so the
detectors fire when there is a match to one of these view-tuned representations.
These representations are probably supported by tissue in the infero­
temporal cortex, near the terminus of the what pathway (see Figure 3.10).
Recording from cells in this area has shown that many neurons here seem
object-specific — that is, they fire preferentially when a certain type of object
is on the scene. (For an example of just how specific these cells can be in their
Descendants of the Feature Net
•
131
FIGURE 4.13
THE JENNIFER ANISTON CELL
Researchers in one study were able to do single-cell recording within the brains of people who were undergoing
surgical treatment for epilepsy. The researchers located cells that fired strongly whenever a picture of Jennifer
Aniston was in view — whether the picture showed her close up (picture 32) or far away (picture 29), with long
hair (picture 32) or shorter (picture 5). These cells are largely viewpoint-independent; other cells, though, are
viewpoint-dependent.
TEST YOURSELF
9. H
ow does the
McClelland and
Rumelhart model
differ from the older,
“classical” version of
the feature net?
10. On what issues is
there disagreement
between the recognition by components
(RBC) proposal and
the recognition via
multiple views proposal? On what issues
is there agreement?
132 •
“preferred” target, see Figure 4.13.) Crucially, though, many of these neurons
are view-tuned: They fire most strongly to a particular view of the target
object. This is just what one might expect with the multiple-views proposal
(Peissig & Tarr, 2007).
However, there has been lively debate between advocates of the RBC
approach (with its claim that recognition is largely viewpoint-independent)
and the multiple-views approach (with its argument that recognition is
viewpoint-dependent). And this may be a case in which both sides are right —
with some brain tissue being sensitive to viewpoint, and some brain tissue
not being sensitive (see Figure 4.14). Moreover, the perceiver’s task may be
crucial. Some neuroscience data suggest that categorization tasks (“Is this
a cup?”) may rely on viewpoint-independent processing in the brain, while
identification tasks (“Is this the cup I showed you before?”) may rely on
viewpoint-dependent processing (Milivojevic, 2012).
In addition, other approaches to object recognition are being explored (e.g.,
Hayward, 2012; Hummel, 2013; Peissig & Tarr, 2007; Ullman, 2007). Obviously,
there is disagreement in this domain. Even so, let’s be clear that all of the avail­
able proposals involve the sort of hierarchical network we’ve been discussing.
In other words, no matter how the debate about object recognition turns out, it
looks like we’re going to need a network model along the lines we’ve considered.
C H A P T E R F O U R Recognizing Objects
1.4
1.4
1.2
1.2
1.0
0.8
0.6
0.4
0.2
0
A
Right fusiform area
BOLD signal change (%)
BOLD signal change (%)
Left fusiform area
Repeated objects
0.8
0.6
0.4
0.2
0
New Same Different
objects view
view
B
FIGURE 4.14 VIEWPOINT
INDEPENDENCE
1.0
New Same Different
objects view
view
Repeated objects
Is object recognition viewpoint-dependent? Some
aspects of object recognition may be viewpointdependent while other aspects are not. Here, researchers documented viewpoint independence in
the left occipital cortex (Panel A), and so the activity in the fusiform area was the same even when
an object was viewed from a novel perspective.
However, as we see in Panel B, other data show
viewpoint dependence in the right occipital cortex.
Face Recognition
We began our discussion of network models with a focus on how people rec­
ognize letters and words. We’ve now extended our reach and considered how
a network might support the recognition of three-dimensional objects. But
there’s one type of recognition that seems to demand a different approach:
the recognition of faces.
Faces Are Special
As we described at the start of this chapter, damage to the visual system
can produce a disorder known as agnosia — an inability to recognize cer­
tain stimuli — and one type of agnosia specifically involves the perception
of faces. People who suffer from prosopagnosia generally have normal
vision. Indeed, they can look at a photograph and correctly say whether
the photo shows a face or something else; they can generally say whether
a face is a man’s or a woman’s, and whether it belongs to someone young
or someone old. But they can’t recognize individual faces — not even of
their own parents or children, whether from photographs or “live.” They
can’t recognize the faces of famous performers or politicians. In fact, they
can’t recognize themselves (and so they sometimes think they’re looking
through a window at a stranger when they’re actually looking at them­
selves in a mirror).
Often, this condition is the result of brain damage, but in some people
it appears to be present from birth, without any detectable brain damage
(e.g., Duchaine & Nakayama, 2006). Whatever its origin, prosopagnosia
seems to imply the existence of special neural structures involved almost
exclusively in the recognition and discrimination of faces. Presumably,
Face Recognition
•
133
prosopagnosia results from some problem or limitation in the functioning
of this brain tissue. (See Behrman & Avidan, 2005; Burton, Young, Bruce,
Johnston, & Ellis, 1991; Busigny, Graf, Mayer, & Rossion, 2010; Damasio,
Tranel, & Damasio, 1990; De Renzi, Faglioni, Grossi, & Nichelli, 1991. For
a related condition, involving an inability to recognize voices, see Shilowich
& Biederman, 2016.)
The special nature of face recognition is also suggested by a pattern that is
the opposite of prosopagnosia. Some people seem to be “super-recognizers”
and are magnificently accurate in face recognition, even though they have
no special advantage in other perceptual or memory tasks (e.g., Bobak,
Hancock, & Bate, 2015; Davis, Lander, Evans, & Jansari, 2016; Russell,
Duchaine, & Nakayama, 2009; Tree, Horry, Riley, & Wilmer, 2017). These
people are consistently able to remember (and recognize) faces that they
viewed only briefly at some distant point in the past, and they’re also more
successful in tasks that require “face matching” — that is, judging whether
two different views of a face actually show the same person.
There are certainly advantages to being a super-recognizer, but also some
disadvantages. On the plus side, being able to remember faces is obviously
a benefit for a politician or a sales person; super-recognizers also seem to
be much more accurate as eyewitnesses (e.g., in selecting a culprit from a
police lineup). In fact, London’s police force now has a special unit of superrecognizers involved in many aspects of crime investigation (Keefe, 2016).
On the downside, being a super-recognizer can produce some social awk­
wardness. Imagine approaching someone and cheerfully announcing, “I
know you! You used to work at the grocery store on Main Street.” The other
person (who, let’s say, did work in that grocery eight years earlier) might find
this puzzling, perhaps creepy, and maybe even alarming.
What about the rest of us — people who are neither prosopagnosic nor
super-recognizers? It turns out that people differ widely in their ability to
remember and recognize faces (Bindemann, Brown, Koyas, & Russ, 2012;
DeGutis, Wilmer, Mercado, & Cohan, 2013; Wilmer, 2017). These differ­
ences, from person to person, are easy to measure, and there are online
face memory tests that can help you find out whether you’re someone who
has trouble recognizing faces. (If you’re curious, point your browser at the
Cambridge Face Memory Test.)
In all people, though, face recognition seems to involve processes dif­
ferent from those used for other forms of recognition. For example, we’ve
mentioned the debate about whether recognition of houses, or teacups,
or automobiles is viewpoint-dependent. There is no question about this
issue, however, when we’re considering faces: Face recognition is strongly
dependent on orientation, and so it shows a powerful inversion effect. In
one study, four categories of stimuli were considered — right-side-up faces,
upside-down faces, right-side-up pictures of common objects other than
faces, and upside-down pictures of common objects. As Figure 4.15 shows,
performance suffered for all of the upside-down (i.e., inverted) stimuli.
However, this effect was much larger for faces than for other kinds of
134 •
C H A P T E R F O U R Recognizing Objects
Number of errors
6
Faces
4
Houses
FIGURE 4.15
2
Upright
Inverted
FACES AND THE INVERSION EFFECT
People’s memory for faces is quite good, when compared with memory for
other pictures (in this case, pictures of houses). However, performance is
very much disrupted when the pictures of faces are inverted. Performance
with houses is also worse with inverted pictures, but the effect of inversion
is much smaller.
( after yin , 1969)
stimuli (Bruyer, 2001; Yin, 1969). Moreover, with non-faces, the (relatively
small) effect of inversion becomes even smaller with practice; with faces,
the effect of inversion remains in place even after practice (McKone,
Kanwisher, & Duchaine, 2007).
The role of orientation in face recognition can also be illustrated infor­
mally. Figure 4.16 shows two upside-down photographs of former British
prime minister Margaret Thatcher (from Thompson, 1980). You can prob­
ably tell that something is odd about them, but now try turning the book
upside down so that the faces are right side up. As you can see, the difference
FIGURE 4.16 PERCEPTION
OF UPSIDE-DOWN FACES
The left-hand picture looks somewhat odd, but the two pictures
still look relatively similar to each
other. Now, try turning the book
upside down (so that the faces are
upright). In this position, the lefthand face (now on the right) looks
ghoulish, and the two pictures look
very different from each other. Our
perception of upside-down faces is
apparently quite different from our
perception of upright faces.
( after thompson , 1980)
Face Recognition
•
135
between the faces is striking, and yet this fiendish contrast is largely lost
when the faces are upside down. (Also see Rhodes, Brake, & Atkinson, 1993;
Valentine, 1988.)
Plainly, then, face recognition is strongly dependent on orientation in
ways that other forms of object recognition are not. Once again, though, we
need to acknowledge an ongoing debate. According to some authors, the rec­
ognition of faces really is in a category by itself, distinct from all other forms
of recognition (e.g., Kanwisher, McDermott, & Chun, 1997). Other authors,
however, offer a different perspective: They agree that face recognition is
special but argue that certain other types of recognition, in addition to faces,
are special in the same way. As one line of evidence, they argue that proso­
pagnosia isn’t just a disorder of face recognition. In one case, for example, a
prosopagnosic bird-watcher lost not only the ability to recognize faces but
also the ability to distinguish the different types of warblers (Bornstein, 1963;
Bornstein, Sroka, & Munitz, 1969). Another patient with prosopagnosia lost
the ability to tell cars apart; she can locate her car in a parking lot only by
reading all the license plates until she finds her own (Damasio, Damasio, &
Van Hoesen, 1982).
Likewise, in Chapter 2, we mentioned neuroimaging data showing that a
particular brain site — the fusiform face area (FFA) — is specifically respon­
sive to faces. (See, e.g., Kanwisher & Yovel, 2006. For a description of other
brain areas involved in face recognition, see Gainotti & Marra, 2011.) One
study, however, suggests that tasks requiring subtle distinctions among birds,
or among cars, can also produce high levels of activation in this brain area
(Gauthier, Skudlarski, Gore, & Anderson, 2000; also Bukach, Gauthier, &
Tarr, 2006). This finding suggests that the neural tissue “specialized” for faces
isn’t used only for faces. (For more on this debate, see, on the one side,
Grill-Spector, Knouf, & Kanwisher, 2004; McKone et al., 2007; Weiner &
Grill-Spector, 2013. On the other side, see McGugin, Gatenby, Gore, &
Gauthier, 2012; Richler & Gauthier, 2014; Stein, Reeder, & Peeler, 2016;
Wallis, 2013; Zhao, Bülthoff, & Bülthoff, 2016.)
What should we make of all this? There’s no question that humans have a
specialized recognition system that’s crucial for face recognition. This system
certainly involves the FFA in the brain, and damage to this system can cause
prosopagnosia. What’s controversial is how exactly we should describe this
system. According to some authors, the system is truly a face recognition
system and will be used for other stimuli only if those stimuli happen to be
“face-like” (see Kanwisher & Yovel, 2006). According to other authors, this
specialized system needs to be defined more broadly: It is used whenever you
are trying to recognize specific individuals within a highly familiar category
(e.g., Gauthier et al., 2000). The recognition of faces certainly has these traits
(e.g., you distinguish Fred from George from Jacob within the familiar category
of “faces”), but other forms of recognition may have the same traits (e.g., if
a bird-watcher is distinguishing different types within the familiar category
of “warblers”).
136 •
C H A P T E R F O U R Recognizing Objects
So far, the data don’t provide a clear resolution of this debate; both sides
of the argument have powerful evidence supporting their view. But let’s focus
on the key point of agreement: Face recognition is achieved by a process
that’s different from the process described earlier in this chapter. We need to
ask, therefore, how face recognition proceeds.
Holistic Recognition
The networks we’ve been considering so far all begin with an analysis of
a pattern’s parts (e.g., features, geons); the networks then assemble those
parts into larger wholes. Face recognition, in contrast, seems not to depend
on an inventory of a face’s parts; instead, this process seems to depend on
holistic perception of the face. In other words, face recognition depends on
the face’s overall configuration — the spacing of the eyes relative to the length
of the nose, the height of the forehead relative to the width of the face, and
so on. (For more on face recognition, see Bruce & Young, 1986; Duchaine &
Nakayama, 2006; Hayward, Crookes, Chu, Favelle, & Rhodes, 2016.)
Of course, a face’s features still matter in this holistic process. The key,
however, is that the features can’t be considered one by one, apart from
the context of the face. Instead, the features matter because of the rela­
tionships they create. It’s the relationships, not the features on their own,
that guide face recognition. (See Fitousi, 2013; Rakover, 2013; Rhodes,
BRAIN AREAS CRUCIAL FOR FACE PERCEPTION
Several brain sites seem to be especially activated when people are looking at faces.
These sites include the fusiform face area (FFA), the occipital face area (OFA), and
the superior temporal sulcus (fSTS).
Face Recognition
•
137
2012; Wang, Li, Fang, Tian, & Liu, 2012, but also see Richler & Gauthier,
2014. For more on holistic perception of facial movement, see Zhao &
Bülthoff, 2017.)
Some of the evidence for this holistic processing comes from the composite effect in face recognition. In an early demonstration of this effect,
Young, Hellawell, and Hay (1987) combined the top half of one face with
the bottom half of another, and participants were asked to identify just
the top half. This task is difficult if the two halves are properly aligned.
In this setting, participants seemed unable to focus only on the top half;
instead, they saw the top of the face as part of the whole (see Figure 4.17A).
Thus, in the figure, it’s difficult to see that the top half of the face is Hugh
Jackman (shown in normal view in Figure 4.17C). This task is relatively easy,
though, if the halves are misaligned (as in Figure 4.17B). Now, the stimulus
itself breaks up the configuration, making it possible to view the top half
on its own. (For related results, see Amishav & Kimchi, 2010; but also see
Murphy, Gray, & Cook, 2017. For evidence that the strength of holistic
processing is predictive of face-recognition accuracy, see Richler, Cheung,
& Gauthier, 2011. For a complication, though, see Rezlescu, Susilo, Wilmer,
& Caramazza, 2017.)
More work is needed to specify how the brain detects and interprets the
relationships that define each face. Also, our theorizing will need to take
some complications into account — including the fact that the recognition
processes used for familiar faces may be different from the processes used
for faces you’ve seen only once or twice (Burton, Jenkins, & Schweinberg,
2011; Burton, Schweinberger, Jenkins, & Kaufmann, 2015; Young & Burton,
2017). Evidence suggests that in recognizing familiar faces, you rely more
heavily on the relationships among the internal features of the face; for unfa­
miliar faces, you may be more influenced by the face’s outer parts such as the
hair and the overall shape of the head (Campbell et al., 1999).
Moreover, psychologists have known for years that people are more
accurate in recognizing faces of people from their own racial background
(e.g., Caucasians looking at other Caucasians, or Asians looking at other
Asians) than they are when trying to recognize people of other races (e.g.,
Meissner & Brigham, 2001). In fact, some people seem entirely prosop­
agnosic when viewing faces of people from other groups, even though
they have no difficulty recognizing faces of people from their own group
(Wan et al., 2017). These points may suggest that people rely on different
mechanisms for, say, “same-race” and “cross-race” face perception, and this
point, too, must be accommodated in our theorizing. (For recent discus­
sions, see Horry, Cheong, & Brewer, 2015; Wan, Crookes, Reynolds, Irons,
& McKone, 2015.)
Obviously, there is still work to do in explaining how we recognize our
friends and family — not to mention how we manage to remember and recog­
nize someone we’ve seen only once before. We know that face recognition re­
lies on processes different from those discussed earlier in the chapter, and we
know that these processes rely on the configuration of the face, rather than
138 •
C H A P T E R F O U R Recognizing Objects
FIGURE 4.17
THE COMPOSITE EFFECT IN FACE RECOGNITION
A
B
C
D
Participants were asked to identify the top half of composite faces like those in
Panels A and B. This task was much harder if the halves were properly aligned
(as in Panel A), and easier if the halves weren’t aligned (as in Panel B). With
the aligned faces, participants have a difficult time focusing on just the face’s
top (and so have a hard time recognizing Hugh Jackman — shown in Panel C).
Instead, they view the face as a whole, and this context changes their perception of Jackman’s features, making it harder to recognize him. (The bottom of
the composite face belongs to Justin Timberlake, shown in Panel D.)
Face Recognition
•
139
TEST YOURSELF
11. W
hat’s the evidence
that face recognition is different from
other forms of object
recognition?
12. What’s the evidence
that face recognition
depends on the face’s
configuration, rather
than the features one
by one?
its individual features. More research is needed, though, to fill in the details of
this holistic processing. (For examples of other research on memory for faces,
see Jones & Bartlett, 2009; Kanwisher, 2006; Michel, Rossion, Han, Chung,
& Caldara, 2006; Rhodes, 2012. For discussion of how these issues play out
in the justice system, with evidence coming from eyewitness identifications,
see Reisberg, 2014.)
Top-Down Influences on Object
Recognition
We’ve now discussed one important limitation of feature nets. These nets
can, as we’ve seen, accomplish a great deal, and they’re crucial for the rec­
ognition of print, three-dimensional objects in the visual environment, and
probably sounds as well. But there are some targets — faces, and perhaps
others — for which recognition depends on configurations rather than indi­
vidual features.
It turns out, though, that there is another limit on feature nets, even if
we’re focusing on the targets for which a feature net is useful — print, com­
mon objects, and so on. Even in this domain, feature nets must be supple­
mented with additional mechanisms. This requirement doesn’t undermine the
importance of the feature net idea; feature nets are definitely needed as part
of our theoretical account. The key word, however, is “part,” because we
need to place feature nets within a larger theoretical frame.
The Benefits of Larger Contexts
Earlier in the chapter, we saw that letter recognition is improved by context.
For example, the letter V is easier to recognize in the context “VASE,” or
even the nonsense context “VIMP,” than it is if presented alone. These are
examples of “top-down” effects — effects driven by your knowledge and ex­
pectations. And these particular top-down effects, based on spelling patterns,
are easily accommodated by the network: As we have discussed, priming
(from recency and frequency of use) guarantees that detectors that have often
been used in the past will be easier to activate in the future. In this way, the
network “learns” which patterns are common and which are not, and it is
more receptive to inputs that follow the usual patterns.
Other top-down effects, however, require a different type of explanation.
Consider the fact that words are easier to recognize if you see them as part
of a sentence than if you see them in isolation. There have been many formal
demonstrations of this effect (e.g., Rueckl & Oden, 1986; Spellman, Holyoak,
& Morrison, 2001; Tulving & Gold, 1963; Tulving, Mandler, & Baumal,
1964), but for our purposes an informal example will work. Imagine that we
tell research participants, “I’m about to show you a word very briefly on a
computer screen; the word is the name of something that you can eat.” If we
forced the participants to guess the word at this point, they would be unlikely
140 •
C H A P T E R F O U R Recognizing Objects
to name the target word. (There are, after all, many things you can eat, so
the chances are slim of guessing just the right one.) But if we briefly show
the word “CELERY,” we’re likely to observe a large priming effect; that is,
participants are more likely to recognize “CELERY” with this cue than they
would have been without the cue.
Think about what this priming involves. First, the person needs to under­
stand each of the words in the instruction. If she didn’t understand the word
“eat” (e.g., if she mistakenly thought we had said, “something that you can
beat”), we wouldn’t get the priming. Second, the person must understand the
relations among the words in the instruction. For example, if she mistakenly
thought we had said, “something that can eat you,” we would expect a very
different sort of priming. Third, the person has to know some facts about the
world — namely, the kinds of things that can be eaten; without this knowl­
edge, we would expect no priming.
Obviously, then, this instance of priming relies on a broad range of
knowledge, and there is nothing special about this example. We could ob­
serve similar priming effects if we tell someone that the word about to be
shown is the name of a historical figure or that the word is related to the Star
Wars movies. In each case, the instruction would facilitate perception, with
the implication that in order to explain these various priming effects, we’ll
need to hook up our object-recognition system to a much broader library of
information.
Here’s a different example, this time involving what you hear. Participants
in one study listened to a low-quality recording of a conversation. Some par­
ticipants were told they were listening to an interview with a job candidate;
others were told they were listening to an interview with a suspect in a crimi­
nal case (Lange, Thomas, Dana, & Dawes, 2011). This difference in context
had a powerful effect on what the participants heard. For example, the audio
contained the sentence “I got scared when I saw what it’d done to him.”
Participants who thought they were listening to a criminal often mis-heard
this statement and were sure they had heard “. . . when I saw what I’d done
to him.”
Where does all of this bring us? Examples like we’re considering here tell
us that we cannot view object recognition as a self-contained process. Instead,
knowledge that is external to object recognition (e.g., knowledge about what
is edible, or about the sorts of things a criminal might say) is imported into
and influences the process. In other words, these examples (unlike the ones
we considered earlier in the chapter) don’t depend just on the specific stimuli
you’ve encountered recently or frequently. Instead, what’s crucial for this
sort of priming is what you know coming into the experiment — knowledge
derived from a wide range of life experiences.
We have, therefore, reached an important juncture. We’ve tried in this
chapter to examine object recognition apart from other cognitive processes,
considering how a separate object-recognition module might function, with
the module then handing its product (the object it had recognized) on to
subsequent processes. We have described how a significant piece of object
TEST YOURSELF
13. W
hat’s the evidence
that word recognition
(or object recognition
in general) is influenced
by processes separate from what has
been seen recently or
frequently?
Top-Down Influences on Object Recognition
•
141
FIGURE 4.18
THE FLOW OF TOP-DOWN PROCESSING
LH Orbitofrontal Cortex
RH Fusiform
LH Fusiform
LH
10–4
RH
p<
10–2
130 ms
180 ms
215 ms
When viewers had only a very brief glimpse of a target object, brain activity
indicating top-down processing was evident in the front part of the brain (the
orbitofrontal cortex) 130 ms after the target came into view. Roughly 50 ms
later (and so 180 ms after the target came into view), brain activity increased
further back in the brain (in the right hemisphere’s fusiform area), indicating
successful recognition. This pattern was not evident when object recognition
was easy (because of a longer presentation of the target). Sensibly, top-down
processing plays a larger role when bottom-up processing is somehow limited
or inadequate.
recognition might proceed, but in the end we have run up against a problem —
namely, top-down priming that draws on knowledge from outside of object
recognition itself. (For neuroscience evidence that word and object recogni­
tion interacts with other sorts of information, see Carreiras et al., 2014; also
Figure 4.18.) This sort of priming depends on what is in memory and on how
that knowledge is accessed and used, and so we can’t tackle this sort of prim­
ing until we’ve said more about memory, knowledge, and thought. We there­
fore must leave object recognition for now in order to fill in other pieces of
the puzzle. We’ll have more to say about object recognition in later chapters,
once we have some additional theoretical machinery in place.
COGNITIVE PSYCHOLOGY AND EDUCATION
speed-reading
Students often wish they could read more quickly, and, in fact, it’s easy to
teach people how to speed-read. It’s important to understand, however, how
speed-reading works, because this will help you see when speed-reading is a
good idea — and when it’s a terrible strategy.
142 •
C H A P T E R F O U R Recognizing Objects
As the chapter describes, in normal reading there’s no need to look at every
word on the page. Printed material (like language in general) follows predictable
patterns, and so, having read a few words, you’re often able to guess what the
next words will be. And without realizing you’re doing it, you’re already exploit­
ing this predictability. In reading this (or any) page, your eyes skip over many of
the words, and you rely on rapid inference to fill in what you’ve skipped.
The same process is central for speed-reading. Courses that teach you
how to speed-read actually encourage you to skip more, as you move down
the page, and to rely more on inference. As a result, speed-reading isn’t really
“reading faster”; it is instead “reading less and inferring more.”
How does this process work? First, before you speed-read some text, you
need to lay the groundwork for the inference process — so that you’ll make
the inferences efficiently and accurately. Therefore, before you speed-read a
text, you should flip through it quickly. Look at the figures and the figure
captions. If there’s a summary at the end or a preview at the beginning, read
them. These steps will give you a broad sense of what the material is about,
preparing you to make rapid — and sensible — inferences about the material.
Second, you need to make sure you do rely on inference, rather than
word-by-word scrutiny of the page. To do this, read for a while holding an
index card just under the line you’re reading, or using your finger to slide
along the line of print to indicate what you’re reading at that moment. These
WHEN SHOULD YOU SPEED-READ?
Students are often assigned an enormous amount of reading, so strategies for
speed-reading can be extremely helpful. But it’s important to understand why speedreading works as it does; knowing this will help you decide when speed-reading is
appropriate and when it’s unwise.
Cognitive Psychology and Education
•
143
procedures establish a physical marker that helps you keep track of where
your eyes are pointing as you move from word to word.
This use of a pointer will become easy and automatic after a little practice,
and once it does, you’re ready for the key step. Rather than using the marker
to follow your eye position, use the marker to lead your eyes. Specifically, try
moving the index card or your finger a bit more quickly than you have so far,
and try to move your eyes to “keep up” with this marker.
Of course, if you suddenly realize that you don’t have a clue what’s on the
page, then you’ve been going too fast. Just move quickly enough so that you
have to hustle along to keep up with your pointer. Don’t move so quickly that
you lose track of what you’re reading.
This procedure will feel awkward at first, but it will become easier with
practice, and you’ll gradually learn to move the pointer faster and faster.
As a result, you’ll increase your reading speed by 30%, 40%, or more. But
let’s be clear about what’s going on here: You’re simply shifting the balance
between how much input you’re taking in and how much you’re filling in the
gaps with sophisticated guesswork. Often, this is a fine strategy. Many of the
things you read are highly predictable, so your inferences about the skipped
words are likely to be correct. In settings like these, you might as well use the
faster process of making inferences, rather than the slower process of looking
at individual words.
But speed-reading is a bad bet if the material is hard to understand.
In that case, you won’t be able to figure out the skipped words via infer­
ence, so speed-reading will hurt you. Speed-reading is also a poor choice if
you’re trying to appreciate an author’s style. If, for example, you speed-read
Shakespeare’s Romeo and Juliet, you probably will be able to make infer­
ences about the plot, but you won’t be able to make inferences about the
specific words you’re skipping over; you won’t be able to make inferences
about the language Shakespeare actually used. And, of course, if you miss the
language of Shakespeare and miss the poetry, you’ve missed the point.
Speed-reading will enable you to zoom through many assignments. But
don’t speed-read material that’s technical, filled with details that you’ll need,
or beautiful for its language. In those cases, you need to pay attention to the
words on the page and not rely on your own inferences.
For more on this topic . . .
Rayner, K., Schotter, E.R., Masson, M.E.J., Potter, M.C., & Treiman, R. (2016). So
much to read, so little time: How do we read, and can speed reading help?
Psychological Science in the Public Interest, 17, 4–34.
144 •
C H A P T E R F O U R Recognizing Objects
chapter review
SUMMARY
• We easily recognize a wide range of objects in
a wide range of circumstances. Our recognition is
significantly influenced by context, which can deter­
mine how or whether we recognize an object. To
study these achievements, investigators have often
focused on the recognition of printed language,
using this case to study how object recognition in
general might proceed.
• Many investigators have proposed that recogni­
tion begins with the identification of features in the
input pattern. Key evidence for this claim comes
from neuroscience studies showing that the detection
of features is separate from the processes needed to
assemble these features into more complex wholes.
• To study word recognition, investigators often
use tachistoscopic presentations. In these studies,
words that appear frequently in the language are
easier to identify than words that don’t appear fre­
quently, and so are words that have been recently
viewed — an effect known as repetition priming.
The data also show a pattern known as the “wordsuperiority effect”; this refers to the fact that let­
ters are more readily perceived if they appear in the
context of a word than if they appear in isolation.
In addition, well-formed nonwords are more readily
perceived than letter strings that do not conform to
the rules of normal spelling. Another reliable pat­
tern is that recognition errors, when they occur, are
quite systematic, with the input typically perceived
as being more regular than it actually is. These find­
ings, taken together, indicate that recognition is
influenced by the regularities that exist in our envi­
ronment (e.g., the regularities of spelling patterns).
• We can understand these results in terms of a
network of detectors. Each detector collects input
and fires when the input reaches a threshold level.
A network of these detectors can accomplish a great
deal; for example, it can interpret ambiguous in­
puts, recover from its own errors, and make infer­
ences about barely viewed stimuli.
• The feature net seems to “know” the rules of
spelling and “expects” the input to conform to these
rules. However, this knowledge is distributed across
the entire network and emerges only through the
network’s parallel processing. This setup leads to
enormous efficiency in our interactions with the
world because it enables us to recognize patterns
and objects with relatively little input and under
diverse circumstances. But these gains come at the
cost of occasional error. This trade-off may be nec­
essary, though, if we are to cope with the informa­
tional complexity of our world.
• A feature net can be implemented in different
ways — with or without inhibitory connections, for
example. With some adjustments (e.g., the addition
of geon detectors), the net can also recognize threedimensional objects. However, some stimuli — for
example, faces — probably are not recognized
through a feature net but, instead, require a differ­
ent sort of recognition system, one that is sensitive
to relationships and configurations within the stim­
ulus input.
• The feature net also needs to be supplemented to
accommodate top-down influences on object recog­
nition. These influences can be detected in the bene­
fits of larger contexts in facilitating recognition and
in forms of priming that are concept-driven rather
than data-driven. These other forms of priming call
for an interactive model that merges bottom-up and
top-down processes.
145
KEY TERMS
bottom-up processing (p. 111)
top-down processing (p. 111)
visual search task (p. 111)
integrative agnosia (p. 112)
tachistoscope (p. 113)
mask (p. 113)
priming (p. 114)
repetition priming (p. 114)
word-superiority effect (WSE) (p. 114)
well-formedness (p. 116)
feature nets (p. 118)
activation level (p. 118)
response threshold (p. 118)
recency (p. 119)
frequency (p. 119)
bigram detectors (p. 120)
local representation (p. 126)
distributed representation (p. 126)
excitatory connections (p. 128)
inhibitory connections (p. 128)
recognition by components (RBC) model (p. 129)
geons (p. 130)
viewpoint-independent recognition (p. 131)
viewpoint-dependent recognition (p. 131)
prosopagnosia (p. 133)
inversion effect (p. 134)
holistic perception (p. 137)
TEST YOURSELF AGAIN
1.What is the difference between “bottom-up”
and “top-down” processing?
2.What is the evidence that features play a special
role in object recognition?
3.What is repetition priming, and how is it
demonstrated?
4.What procedure demonstrates the wordsuperiority effect?
5.What’s the evidence that word perception is
somehow governed by the rules of ordinary
spelling?
6.How does a feature net explain the wordfrequency effect?
7.How does a feature net explain the types of
errors people make in recognizing words?
8.What are the benefits, and what are the costs,
associated with the feature net’s functioning?
146
9.How does the McClelland and Rumelhart
model differ from the older, “classical” version
of the feature net?
10.On what issues is there disagreement between
the recognition by components (RBC) proposal
and the recognition via multiple views proposal?
On what issues is there agreement?
11.What’s the evidence that face recognition
is different from other forms of object
recognition?
12.What’s the evidence that face recognition
depends on the face’s configuration, rather
than the features one by one?
13.What’s the evidence that word recognition
(or object recognition in general) is influenced
by processes separate from what has been seen
recently or frequently?
THINK ABOUT IT
1.Imagine that you were designing a mechanism
that would recognize items of clothing (shirts,
pants, jackets, belts). Would some sort of
feature net be possible? If so, what would the
net involve?
2.Imagine that you were designing a mechanism
that would recognize different smells (roses,
cinnamon, freshly mown grass, car exhaust).
Do you think some sort of feature net would be
possible? If so, what would the net involve?
E eBOOK DEMONSTRATIONS & ESSAYS
Go to http://digital.wwnorton.com/cognition7 for the online demonstrations and essays
relevant to this chapter. These can be found in the ebook.
Online Demonstrations
Online Applying Cognitive Psychology and the
Law Essays
• Demonstration 4.1: Features and Feature
• Cognitive Psychology and the Law: Cross-Race
Identification
Combination
• Demonstration 4.2: The Broad Influence of the
Rules of Spelling
• Demonstration 4.3: Inferences in Reading
COGNITION LABS
Go to http://digital.wwnorton.com/cognition7 to experience interactive online psychology
labs relevant to this chapter.
147
5
chapter
Paying Attention
what if…
Right now, you’re paying attention to this page, reading these words. But you could, if you chose, pay
attention to the other people in the room, or your plans for the weekend,
or even the feel of the floor under your feet.
What would your life be like if you couldn’t control your attention in
this way? Every one of us has, of course, had the maddening experience
of being distracted when we’re trying to concentrate. For example, there
you are on the bus, trying to read your book. You have no interest in the
conversation going on in the seats behind you, but you seem unable to
shut it out, and so you make no progress in your book.
The frustration in this experience is surely fueled by the fact that usually
you can control your attention, so it’s especially irritating when you can’t
focus in the way you want to. The life challenge is much worse, though,
for people who suffer from attention deficit disorder. This disorder is often
associated with hyperactivity — and hence the abbreviation ADHD. People with ADHD are often overwhelmed by the flood of information that’s
available to them, and they’re unable to focus on their chosen target.
The diagnosis of ADHD can range from relatively mild to quite severe,
and we’ll have more to say about ADHD later in the chapter. Nothing
in this range, though, approaches a much more extreme disruption in
attention termed “unilateral neglect syndrome.” This pattern is generally the result of damage to the parietal cortex, and patients with this
syndrome ignore all inputs coming from one side of the body. A patient
with neglect syndrome will, for example, eat food from only one side
of the plate, wash only half of his or her face, and fail to locate soughtfor objects if they’re on the neglected side (see Logie, 2012; Sieroff,
Pollatsek, & Posner, 1988). Someone with this disorder cannot safely
drive a car and, as a pedestrian, is likely to trip over unnoticed obstacles.
This syndrome typically results from damage to the right parietal
lobe, and so the neglect is for the left side of space. (Remember the
brain’s contralateral organization; see Chapter 2.) Neglect patients will
therefore read only the right half of words shown to them — they’ll read
“threat” as “eat,” “parties” as “ties.” If asked to draw a clock, they’ll probably remember that the numbers from 1 to 12 need to be included, but
they’ll jam all the numbers into the clock’s right side.
All these observations remind us just how crucial the ability to pay
attention is — so that you can focus on the things you want to focus
149
preview of chapter themes
•
•
•
ultiple mechanisms are involved in the seemingly
M
simple act of paying attention, because people must take
various steps to facilitate the processing of desired inputs.
Without these steps, their ability to pick up information
from the world is dramatically reduced.
•
any of the steps necessary for perception have a “cost”:
M
They require the commitment of mental resources. These
resources are limited in availability, which is part of the reason you usually can’t pay attention to two inputs at once—
doing so would require more resources than you have.
ome of the mental resources you use are specialized,
S
which means they’re required only for tasks of a certain
sort. Other resources are more general, needed for a wide
range of tasks. However, the resource demand of a task
can be diminished through practice.
•
e emphasize that attention is best understood not as a
W
process or mechanism but as an achievement. Like most
achievements, paying attention involves many elements, all
of which help you to be aware of the stimuli you’re interested in and not be pulled off track by irrelevant distractors.
perform two activities at the same time only if the activities don’t require more resources than you have available.
ivided attention (the attempt to do two things at once)
D
can also be understood in terms of resources. You can
on and not be pulled off track by distraction. But what is “attention”?
As we’ll see in this chapter, the ability to pay attention involves many
independent elements.
Selective Attention
William James (1842–1910) is one of the historical giants of the field of
psychology, and he is often quoted in the modern literature. One of his most
famous quotes provides a starting point for this chapter. Roughly 125 years
ago, James wrote:
Everyone knows what attention is. It is the taking possession by the
mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought. Focalization, concentration,
of consciousness are of its essence. It implies withdrawal from some
things in order to deal effectively with others, and is a condition which
has a real opposite in the confused, dazed, scatterbrained state which in
French is called distraction. . . . (James, 1890, pp. 403–404)
In this quote, James is describing what modern psychologists call selective
attention — that is, the skill through which a person focuses on one input or
one task while ignoring other stimuli that are also on the scene. But what
does this skill involve? What steps do you need to take in order to achieve the
focus that James described, and why is it that the focus “implies withdrawal
from some things in order to deal effectively with others”?
Dichotic Listening
Early studies of attention used a setup called dichotic listening: Participants
wore headphones and heard one input in the left ear and a different input
in the right ear. The participants were instructed to pay attention to one of
these inputs — the attended channel — and to ignore the message in the other
ear — the unattended channel.
150 •
C H A P T E R F I V E Paying Attention
FIGURE 5.1
THE INVISIBLE GORILLA
In this procedure, participants are instructed to keep track of the ballplayers in the white shirts. Intent on their
task, participants are oblivious to what the black-shirted players are doing, and—remarkably—they fail to see the
person in the gorilla suit strolling through the scene.
(figure provided by daniel j. simons.)
To make sure participants were paying attention, investigators gave them a
task called shadowing: Participants were required to repeat back what they were
hearing, word for word, so that they were echoing the attended channel. Their
shadowing performance was generally close to perfect, and they were able to
echo almost 100% of what they heard. At the same time, they heard remarkably
little from the unattended channel. If asked, after a minute or so of shadowing,
to report what the unattended message was about, they had no idea (e.g., Cherry,
1953). They couldn’t even tell if the unattended channel contained a coherent
message or random words. In fact, in one study, participants shadowed speech in
the attended channel, while in the unattended channel they heard a text in Czech,
read with English pronunciation. The individual sounds, therefore (the vowels,
the consonants), resembled English, but the message itself was (for an English
speaker) gibberish. After a minute of shadowing, only 4 of 30 participants
detected the peculiar character of the unattended message (Treisman, 1964).
We can observe a similar pattern with visual inputs. Participants in one study
viewed a video that has now gone viral on the Internet and is widely known
as the “invisible gorilla” video. In this video, a team of players in white shirts
is passing a basketball back and forth; people watching the video are urged to
count how many times the ball is passed from one player to another. Interwoven
with these players (and visible in the video) is another team, wearing black shirts,
also passing a ball back and forth; viewers are instructed to ignore these players.
Viewers have no difficulty with this task, but, while doing it, they usually don’t see another event that appears on the screen right in front of their
eyes. Specifically, they fail to notice when someone wearing a gorilla costume
walks through the middle of the game, pausing briefly to thump his chest
before exiting. (See Figure 5.1; Neisser & Becklen, 1975; Simons & Chabris,
1999; also see Jenkins, Lavie, & Driver, 2005.)
Selective Attention
•
151
Even so, people are not altogether oblivious to the unattended channel.
In selective listening experiments, research participants easily and accurately
report whether the unattended channel contained human speech, musical
instruments, or silence. If the unattended channel did contain speech, participants can report whether the speaker was male or female, had a high or
low voice, or was speaking loudly or softly. (For reviews of this early work,
see Broadbent, 1958; Kahneman, 1973.) Apparently, then, physical attributes
of the unattended channel are heard, even though participants are generally
clueless about the unattended channel’s semantic content.
In one study, however, participants were asked to shadow one passage
while ignoring a second passage. Embedded within the unattended channel
was a series of names, and roughly one third of the participants did hear their
own name when it was spoken — even though (just like in other studies) they
heard almost nothing else from the unattended input (Moray, 1959).
And it’s not just names that can “catch” your attention. Mention of a recently
seen movie, or of a favorite restaurant, will often be noticed in the unattended
channel. More broadly, words with some personal importance are often noticed,
even though the rest of the unattended channel is perceived only as an undifferentiated blur (Conway, Cowan, & Bunting, 2001; Wood & Cowan, 1995).
Inhibiting Distractors
THE COCKTAIL
PARTY EFFECT
We have all experienced some
version of the so-called cocktail party effect. There you are
at a party, deep in conversation. Other conversations are
going on, but somehow you’re
able to “tune them out.” All
you hear is the single conversation you’re attending to, plus
a buzz of background noise.
But now imagine that someone a few steps away from
you mentions the name of a
close friend of yours. Your attention is immediately caught,
and you find yourself listening to that other conversation
and (momentarily) oblivious
to the conversation you had
been engaged in. This experience, easily observed outside
the laboratory, matches the
pattern of experimental data.
152 •
How can we put all these research results together? How can we explain
both the general insensitivity to the unattended channel and also the cases in
which the unattended channel “leaks through”?
One option focuses on what you do with the unattended input. The proposal is that you somehow block processing of the inputs you’re not interested in, much as a sentry blocks the path of unwanted guests but stands
back and does nothing when legitimate guests are in view, allowing them
to pass through the gate unimpeded. This sort of proposal was central for
early theories of attention, which suggested that people erect a filter that
shields them from potential distractors. Desired information (the attended
channel) is not filtered out and so goes on to receive further processing
(Broadbent, 1958).
But what does it mean to “filter” something out? The key lies in the nervous system’s ability to inhibit certain responses, and evidence suggests that
you do rely on this ability to avoid certain forms of distraction. This inhibition, however, is rather specific, operating on a distractor-by-distractor basis.
In other words, you might have the ability to inhibit your response to this
distractor and the same for that distractor, but these abilities are of little
value if some new, unexpected distractor comes along. In that case, you need
to develop a new skill aimed at blocking the new intruder. (See Cunningham
& Egeth, 2016; Fenske, Raymond, Kessler, Westoby, & Tipper, 2005; Frings
& Wühr, 2014; Jacoby, Lindsay, & Hessels, 2003; Tsushima, Sasaki, &
Watanabe, 2006; Wyatt & Machado, 2013. For a glimpse of brain mechanisms
that support this inhibition, see Payne & Sekuler, 2014.)
C H A P T E R F I V E Paying Attention
The ability to ignore certain distractors — to shut them out — therefore
needs to be part of our theory. Other evidence, though, indicates that this
isn’t the whole story. That’s because you not only inhibit the processing of
distractors, you also promote the processing of desired stimuli.
Inattentional Blindness
We saw in Chapters 3 and 4 that perception involves a lot of activity, as
you organize and interpret the incoming stimulus information. It seems plausible that this activity would require some initiative and some resources from
you — and evidence suggests that it does.
In one experiment, participants were told that they would see large “+”
shapes on a computer screen, presented for 200 ms (milliseconds), followed by a
pattern mask. (The mask was just a meaningless jumble on the screen, designed
to disrupt any further processing.) If the horizontal bar of the “+” was longer
than the vertical, participants were supposed to press one button; if the vertical
bar was longer, they had to press a different button. As a complication, participants weren’t allowed to look directly at the “+.” Instead, they fixated on (i.e.,
pointed their eyes at) a mark in the center of the computer screen — a fixation
target — and the “+” shapes were shown just off to one side (see Figure 5.2).
FIGURE 5.2
INATTENTIONAL BLINDNESS
100
90
Percent failing
to see the change
80
70
60
50
40
30
20
10
0
No warning
Warning
Condition
Participants were instructed to point their eyes at the dot and to make judgments about the “+” shown just off to
the side. However, the dot itself briefly changed to another shape. If participants weren’t warned about this (and
so weren’t paying attention to the dot), they failed to detect this change — even though they had been pointing
their eyes right at the dot the whole time.
(after mack & rock, 1998)
Selective Attention
•
153
INATTENTIONAL
BLINDNESS OUTSIDE
THE LAB
Inattentional blindness is usually demonstrated in the laboratory, but it has a number of
real-world counterparts. Most
people, for example, have experienced the peculiar situation in which they can’t find
the mayonnaise in the refrigerator (or the ketchup or the
salad dressing) even though
they’re staring right at the
bottle. This happens because
they’re so absorbed in other
thoughts that they become
blind to an otherwise salient
stimulus.
154 •
For the first three trials of the procedure, events proceeded just as the participants
expected, and the task was relatively easy. On Trial 4, though, things were slightly
different: While the target “+” was on the screen, the fixation target disappeared and
was replaced by one of three shapes — a triangle, a rectangle, or a cross. Then, the
entire configuration (the “+” target and this new shape) was replaced by the mask.
Immediately after the trial, participants were asked: Was there anything different on this trial? Was anything present, or anything changed, that wasn’t there
on previous trials? Remarkably, 89% of the participants reported that there was
no change; they had failed to see anything other than the (attended) “+.” To
probe the participants further, the researchers told them (correctly) that during
the previous trial the fixation target had momentarily disappeared and had been
replaced by a shape. The participants were then asked what that shape had been,
and were given the choices of a triangle, a rectangle, or a cross (one of which,
of course, was the right answer). The responses to this question were essentially
random. Even when probed in this way, participants seemed not to have seen the
shape directly in front of their eyes (Mack & Rock, 1998; also see Mack, 2003).
This pattern has been named inattentional blindness (Mack & Rock,
1998; also Mack, 2003) — a pattern in which people fail to see a prominent
stimulus, even though they’re staring straight at it. In a similar effect, called
“inattentional deafness,” participants regularly fail to hear prominent stimuli
if they aren’t expecting them (Dalton & Fraenkel, 2012). In other studies,
participants fail to feel stimuli if the inputs are unexpected; this is “inattentional numbness” (Murphy & Dalton, 2016).
What’s going on here? Are participants truly blind (or deaf or numb) in
response to these various inputs? As an alternative, some researchers propose
that participants in these experiments did see (or hear or feel) the targets but, a
moment later, couldn’t remember what they’d just experienced (e.g., Wolfe, 1999;
also Schnuerch, Kreiz, Gibbons, & Memmert, 2016). For purposes of theory, this
distinction is crucial, but for now let’s emphasize what the two proposals have in
common: By either account, your normal ability to see what’s around you, and to
make use of what you see, is dramatically dim­inished in the absence of attention.
Think about how these effects matter outside of the laboratory. Chabris
and Simons (2010), for example, call attention to reports of traffic accidents
in which a driver says, “I never saw the bicyclist! He came out of nowhere!
But then — suddenly — there he was, right in front of me.” Drew, Võ, and
Wolfe (2013) showed that experienced radiologists often miss obvious ano­
malies in a patient’s CT scan, even when looking right at the anomaly. (For
similar concerns, related to inattentional blindness in eyewitnesses to crimes,
see Jaeger, Levin, & Porter, 2017.) Or, as a more mundane example, you go
to the refrigerator to find the mayonnaise (or the ketchup or the juice) and
don’t see it, even though it’s right in front of you.
In these cases, we lament the neglectful driver and the careless radiologist,
and your inability to find the mayo may cause you to worry that you’re losing
your mind (as well as your condiments). The reality, though, is that these cases
of failing-to-see are entirely normal. Perception requires more than “merely”
having a stimulus in front of your eyes. Perception requires some work.
C H A P T E R F I V E Paying Attention
Change Blindness
The active nature of perception is also evident in studies of change
blindness — observers’ inability to detect changes in scenes they’re looking
directly at. In some experiments, participants are shown pairs of pictures separated by a brief blank interval (e.g., Rensink, O’Regan, & Clark, 1997). The
pictures in each pair are identical except for one aspect — an “extra” engine
shown on the airplane in one picture and not in the other; a man wearing a
hat in one picture but not wearing one in the other; and so on (see Figure 5.3).
FIGURE 5.3
CHANGE BLINDNESS
In some change-blindness demonstrations, participants see one picture, then a second, then the first again, then
the second, and must spot the difference between the two pictures. Here, we’ve displayed the pictures side by
side, rather than putting them in alternation. Can you find the differences? For most people, it takes a surprising
amount of time and effort to locate the differences — even though some of the differences are large. Apparently,
having a stimulus directly in front of your eyes is no guarantee that you will perceive the stimulus.
Selective Attention
•
155
FIGURE 5.4
CHANGE
BLINDNESS
In this video, every time
there was a shift in camera angle, there was a
change in the scene — so
that the woman in the red
sweater abruptly gained
a scarf, the plates that
had been red were suddenly white, and so on.
When viewers watched
the video, though, they
noticed none of these
changes.
Participants know that their task is to detect any changes in the pictures, but
even so, the task is difficult. If the change involves something central to the
scene, participants may need to look back and forth between the pictures as
many as a dozen times before they detect the change. If the change involves
some peripheral aspect of the scene, as many as 25 alternations may be required.
A related pattern can be documented when participants watch videos. In
one study, observers watched a movie of two women having a conversation.
The camera first focused on one woman, then the other, just as it would in an
ordinary TV show or movie. The crucial element of this experiment, though,
was that certain aspects of the scene changed every time the camera angle
changed. For example, from one camera angle, participants could plainly see
the red plates on the table between the women. When the camera shifted to a
different position, though, the plates’ color had changed to white. In another
shift, one of the women gained a prominent scarf that she didn’t have on a
fraction of a second earlier (see Figure 5.4). Most observers, however, noticed
none of these changes (Levin & Simons, 1997; Shore & Klein, 2000; Simons
& Rensink, 2005).
Incredibly, the same pattern can be documented with live (i.e., not
filmed) events. In a remarkable study, an investigator (let’s call him “Leon”)
approached pedestrians on a college campus and asked for directions to a certain building. During the conversation, two men carrying a door approached
and deliberately walked between Leon and the research participant. As a
result, Leon was momentarily hidden (by the door) from the participant’s
view, and in that moment Leon traded places with one of the men carrying
the door. A second later, therefore, Leon was able to walk away, unseen, while
the new fellow (who had been carrying the door) stayed behind and conti­
nued the conversation with the participant.
Roughly half of the participants failed to notice this switch. They conti­
nued the conversation as though nothing had happened — even though Leon
and his replacement were wearing different clothes and had easily distinguishable voices. When asked whether anything odd had happened in this
event, many participants commented only that it was rude that the guys carrying the door had walked right through their conversation. (See Simons
& Ambinder, 2005; Chabris & Simons, 2010; also see Most et al., 2001;
Rensink, 2002; Seegmiller, Watson, & Strayer, 2011. For similar effects with
auditory stimuli, see Gregg & Samuel, 2008; Vitevitch, 2003.)
Early versus Late Selection
It’s clear, then, that people are often oblivious to stimuli directly in front of
their eyes — whether the stimuli are simple displays on a computer screen,
photographs, videos, or real-life events. (Similarly, people are sometimes
oblivious to prominent sounds in the environment.) As we’ve said, though,
there are two ways to think about these results. First, the studies may reveal
genuine limits on perception, so that participants literally don’t see (or hear)
these stimuli; or, second, the studies may reveal limits on memory, so that
156 •
C H A P T E R F I V E Paying Attention
participants do see (or hear) the stimuli but immediately forget what they’ve
just experienced.
Which proposal is correct? One approach to this question hinges on
when the perceiver selects the desired input and (correspondingly) when
the perceiver stops processing the unattended input. According to the early
selection hypothesis, the attended input is privileged from the start, so that
the unattended input receives little analysis and therefore is never perceived.
According to the late selection hypothesis, all inputs receive relatively complete analysis, and selection occurs after the analysis is fini­shed. Perhaps
the selection occurs just before the stimuli reach consciousness, so that we
become aware only of the attended input. Or perhaps the selection occurs
later still — so that all inputs make it (briefly) into consciousness, but then
the selection occurs so that only the attended input is remembered.
Each hypothesis captures part of the truth. On the one side, there are cases in
which people seem unaware of distractors but are influenced by them anyway —
so that the (apparently unnoticed) distractors guide the interpretation of the
attended stimuli (e.g., Moore & Egeth, 1997; see Figure 5.5). This seems to
be a case of late selection: The distractors are perceived (so that they do have
FIGURE 5.5
A
UNCONSCIOUS PERCEPTION
B
C
One study, apparently showing late selection, found that participants perceived (and were influenced) by background stimuli even though the participants did not consciously perceive these stimuli. The participants were
shown a series of images, each containing a pair of horizontal lines; their task was to decide which line was
longer. For the first three trials, the background dots in the display were arranged randomly (Panel A). For the
fourth trial, the dots were arranged as shown in Panel B, roughly reproducing the configuration of the Müller-Lyer
illusion; Panel C shows the standard form of this illusion. Participants in this study didn’t perceive the “fins” consciously, but they were influenced by them — judging the top horizontal line in Panel B to be longer, fully in accord
with the usual misperception of this illusion.
Selective Attention
•
157
TEST YOURSELF
1.What information do
people reliably pick
up from the attended
channel? What do
they pick up from the
unattended channel?
2.How is inattentional
blindness demonstrated? What situations
outside of the laboratory
seem to reflect inattentional blindness?
3.What evidence seems
to confirm early selection? What evidence
seems to confirm late
selection?
an influence) but are selected out before they make it to consciousness. On
the other side, though, we can also find evidence for early selection, with
attended inputs being privileged from the start and distractor stimuli falling
out of the stream of processing at a very early stage. Relevant evidence comes,
for example, from studies that record the brain’s electrical activity in the milli­
seconds after a stimulus has arrived. These studies confirm that the brain
activity for attended inputs is distinguishable from that for unattended inputs
just 80 ms or so after the stimulus presentation — a time interval in which
early sensory processing is still under way (Hillyard, Vogel, & Luck, 1998;
see Figure 5.6).
Other evidence suggests that attention can influence activity levels in the
lateral geniculate nucleus, or LGN (Kastner, Schneider, & Wunderlich, 2006;
McAlonan, Cavanaugh & Wurtz, 2008; Moore & Zirnsak, 2017; Vanduffel,
Tootell, & Orban, 2000). In this case, attention is changing the flow of signals
within the nervous system even before the signals reach the brain. (For more
on how attention influences processing in the visual cortex, see Carrasco,
Ling, & Read, 2004; Carrasco, Penpeci-Talgar, & Eckstein, 2000; McAdams
& Reid, 2005; Reynolds, Pasternak, & Desimone, 2000; also see O’Connor,
Fukui, Pinsk, & Kastner, 2002; Yantis, 2008.)
Selection via Priming
Whether selection is early or late, it’s clear that people often fail to see stimuli
that are directly in front of them, in plain view. But what is the obstacle here?
Why don’t people perceive these stimuli?
In Chapter 4, we proposed that recognition requires a network of
detectors, and we argued that these detectors fire most readily if they’re
suitably primed. In some cases, the priming is produced by your visual
experience — specifically, whether each detector has been used recently or
frequently in the past. But we suggested that priming can also come from
another source: your expectations about what the stimulus will be.
The proposal, then, is that you can literally prepare yourself for perceiving
by priming the relevant detectors. In other words, you somehow reach into
the network and deliberately activate just those detectors that, you believe,
will soon be needed. Then, once primed in this way, those detectors will be
on “high alert” and ready to fire.
Let’s also suppose that this priming isn’t “free.” Instead, you need to spend
some effort or allocate some resources in order to do the priming, and these
resources are in limited supply. As a result, there’s a limit on just how much
priming you can do.
We’ll need to flesh out this proposal in several ways, but even so, we can
already use it to explain some of the findings we’ve already met. Why don’t
participants notice the shapes in the inattentional blindness studies? The
answer lies in the fact that they don’t expect any stimulus to appear, so they
have no reason to prepare for any stimulus. As a result, when the stimulus
is presented, it falls on unprepared (unprimed, unresponsive) detectors. The
158 •
C H A P T E R F I V E Paying Attention
FIGURE 5.6
EVIDENCE FOR EARLY SELECTION
*
*
Right ear
*
Left ear
Time
A
Electrical activity in the brain in the interval just
after stimulus presentation
–2µV
Attended
N1 efect
0
Unattended
+2µV
0
B
80 100
200
Time (ms)
Participants were instructed to pay attention to the targets arriving in one
ear, but to ignore targets in the other ear (Panel A; dots indicate which of the
input signals were actually targets). During this task, researchers monitored
the electrical activity in the participants’ brains, with special focus on a brain
wave termed the “N1” (so-called because the wave reflects a nega­tive voltage
roughly 100 ms after the target). As Panel B shows, the N1 effect was different
for the attended and unattended inputs within 80 ms of the target’s arrival —
indicating that the attended and unattended inputs were processed differently from a very early stage.
(from hillyard et. al. “electric signs of selective
attention in the human brain,” science 182 © 1973 aaas. reprinted with permission.)
detectors therefore don’t respond to the stimulus, so the participants end up
not perceiving it.
What about selective listening? In this case, you’ve been instructed to
ignore the unattended input, so you have no reason to devote any resources
to this input. Hence, the detectors needed for the distractor message
are unprimed, and this makes it difficult to hear the distractor. But why
Selection via Priming
•
159
does attention sometimes “leak,” so that you do hear some aspects of the
unattended input? Think about what will happen if your name is spoken
on the unattended channel. The detectors for this stimulus are already
primed, but this isn’t because at that moment you’re expecting to hear
your name. Instead, the detectors for your name are primed simply because
this is a stimulus you’ve often encountered in the past. Thanks to this
prior exposure, the activation level of these detectors is already high;
you don’t need to prime them further. So they will fire even if your attention is elsewhere.
Two Types of Priming
The idea before us, in short, has three elements. First, perception is vastly
facilitated by the priming of relevant detectors. Second, the priming is
sometimes stimulus-driven — that is, produced by the stimuli you’ve
encountered (recently or frequently) in the past. This is repetition
priming — priming produced by a prior encounter with the stimulus. This
type of priming takes no effort on your part and requires no resources,
and it’s this sort of priming that enables you to hear your name on the
unattended channel. But third, a different sort of priming is also possible.
This priming is expectation-driven and under your control. In this form of
priming, you deliberately prime detectors for inputs you think are upcoming, so that you’re ready for those inputs when they arrive. You don’t do
this priming for inputs you have no interest in, and you can’t do this priming for inputs you can’t anticipate.
Can we test these claims? In a classic series of studies, Posner and Snyder
(1975) gave participants a straightforward task: A pair of letters was shown
on a computer screen, and participants had to decide, as swiftly as they could,
whether the letters were the same or different. So someone might see “AA”
and answer “same” or might see “AB” and answer “different.”
Before each pair, participants saw a warning signal. In the neutral
condition, the warning signal was a plus sign (“+”). This signal notified
participants that the stimuli were about to arrive but provided no other
information. In a different condition, the warning signal was a letter that
actually matched the stimuli to come. So someone might see the warning signal “G” followed by the pair “GG.” In this case, the warning
signal served to prime the participants for the stimuli. In a third condition,
though, the warning signal was misleading. It was again a letter, but a
different letter from the stimuli to come. Participants might see “H”
followed by the pair “GG.” Let’s consider these three conditions neutral,
primed, and misled.
In this simple task, accuracy rates are very high, but Posner and
Snyder also recorded how quickly people responded. By comparing these
response times (RTs) in the primed and neutral conditions, we can ask
what benefit there is from the prime. Likewise, by comparing RTs in the
160 •
C H A P T E R F I V E Paying Attention
misled and neutral conditions, we can ask what cost there is, if any, from
being misled.
Before we turn to the results, there’s a complication: Posner and Snyder
ran this procedure in two different versions. In one version, the warning
signal was an excellent predictor of the upcoming stimuli. For example, if
the warning signal was an A, there was an 80% chance that the upcoming
stimulus pair would contain A’s. In Posner and Snyder’s terms, the warning
signal provided a “high validity” prime. In a different version of the procedure, the warning signal was a poor predictor of the upcoming stimuli. For
example, if the warning signal was an A, there was only a 20% chance that
the upcoming pair would contain A’s. This was the “low validity” condition
(see Table 5.1).
Let’s consider the low-validity condition first, and let’s focus on those
few occasions in which the prime did match the subsequent stimuli. That
is, we’re focusing on 20% of the trials and ignoring the other 80% for the
moment. In this condition, the participant can’t use the prime as a basis
for predicting the stimuli because the prime is a poor indicator of things
to come. Therefore, the prime should not lead to any specific expectations.
Nonetheless, we do expect faster RTs in the primed condition than in the
neutral condition. Why? Thanks to the prime, the relevant detectors have
just fired, so the detectors should still be warmed up. When the target
stimuli arrive, therefore, the detectors should fire more readily, allowing a
faster response.
TABLE 5.1
D ESIGN OF POSNER AND SNYDER’S EXPERIMENT
TYPICAL SEQUENCE
Type of
Trial
Lowvalidity
Condition
Highvalidity
Condition
Warning
Signal
Test
Stimuli
Provides
Repetition
Priming?
Provides
Basis
for Expectation?
Neutral
+
AA
No
No
Primed
G
GG
Yes
No
Misled
H
GG
No
No
Neutral
+
AA
No
No
Primed
G
GG
Yes
Prime leads to
correct expectation
Misled
H
GG
No
Prime leads to
incorrect expectation
In the low-validity condition, misled trials occurred four times as often as primed
trials (80% vs. 20%). Therefore, participants had no reason to trust the primes and,
correspondingly, no reason to generate an expectation based on the primes. In the
high-validity condition, the arrangement was reversed: Now, primed trials occurred
four times as often as misled trials. Therefore, participants had good reason to trust
the primes and good reason to generate an expectation based on the prime.
(after posner & snyder, 1975)
Selection via Priming
•
161
The results bear this out. RTs were reliably faster (by roughly 30 ms) in
the primed condition than in the neutral condition (see Figure 5.7, left side;
the figure shows the differences between conditions). Apparently, detectors
can be primed by mere exposure to a stimulus, even in the absence of expectations, and so this priming is truly stimulus-based.
What about the misled condition? With a low-validity prime, misleading
the participants had no effect: Performance in the misled condition was the
same as performance in the neutral condition. Priming the “wrong” detector,
it seems, takes nothing away from the other detectors — including the detectors actually needed for that trial. This fits with our discussion in Chapter 4:
Each of the various detectors works independently of the others, and so
priming one detector obviously influences the functioning of that specific
detector but neither helps nor hinders the other detectors.
T HE EFFECTS OF PRIMING ON
STIMULUS PROCESSING
30
20
10
High-validity
condition
Benefit of
priming
40
Low-validity
condition
Cost of being
misled
50
Benefit of
priming
0
10
20
30
40
Cost of being
misled
Difference in response times between neutral
condition and experimental conditions (ms)
Cost
Benefit
FIGURE 5.7
50
As one way of assessing the Posner and Snyder (1975) results, we can subtract the response times for the neutral condition from those for the primed
condition; in this way, we measure the benefits of priming. Likewise, we
can subtract the response times for the neutral condition from those for
the misled condition; in this way, we measure the costs of being misled.
In these terms, the low-validity condition shows a small benefit (from repetition priming) but zero cost from being misled. The high-validity condition,
in contrast, shows a larger benefit — but also a substantial cost. The results
shown here reflect trials with a 300 ms interval between the warning signal
and the test stimuli. Results were somewhat different at other intervals.
162 •
C H A P T E R F I V E Paying Attention
Let’s look next at the high-validity primes. In this condition, people
might see, for example, a “J” as the warning signal and then the stimulus
pair “JJ.” Presentation of the prime itself will fire the J-detectors, and this
should, once again, “warm up” these detectors, just as the low-validity
primes did. As a result, we expect a stimulus-driven benefit from the prime.
However, the high-validity primes may also have another influence: Highvalidity primes are excellent predictors of the stimulus to come. Participants are told this at the outset, and they have lots of opportunity to see
that it’s true. High-validity primes will therefore produce a warm-up effect
and also an expectation effect, whereas low-validity primes produce only
the warm-up. On this basis, we should expect the high-validity primes
to help participants more than low-validity primes — and that’s exactly
what the data show (Figure 5.7, right side). The combination of warm-up
and expectations leads to faster responses than warm-up alone. From the
participants’ point of view, it pays to know what the upcoming stimulus
might be.
Explaining the Costs and Benefits
The data make it clear, then, that we need to distinguish two types of primes.
One type is stimulus-based — produced merely by presentation of the priming
stimulus, with no role for expectations. The other type is expectation-based
and is created only when the participant believes the prime allows a prediction of what’s to come.
These types of primes can be distinguished in various ways, including
the biological mechanisms that support them (see Figure 5.8; Corbetta &
Shulman, 2002; Hahn, Ross, & Stein, 2006; but also Moore & Zirnsak,
2017) and also a difference in what they “cost.” Stimulus-based priming
FIGURE 5.8 BIOLOGICAL
MECHANISMS FOR THE TWO
TYPES OF PRIMING
Brain sites shown in black have been
identified in various studies as involved
in expectation-based (sometimes called “goal
directed”) attention; sites shown in blue have
been implicated in “stimulus-driven” attention.
Sites shown in gray have been identified as
involved in both types of attention.
Selection via Priming
•
163
TEST YOURSELF
4.What are the differences between the
way that stimulusbased priming functions and the way that
expectation-based
priming functions?
5.Why is there a “cost”
associated with being
misled by expectationbased priming?
appears to be “free” — we can prime one detector without taking anything
away from other detectors. (We saw this in the low-validity condition, in
the fact that the misled trials led to responses just as fast as those in the neutral trials.) Expectation-based priming, in contrast, does have a cost, and
we see this in an aspect of Figure 5.7 that we’ve not yet mentioned: With
high-validity primes, responses in the misled condition were slower than
responses in the neutral condition. That is, misleading the participants actually hurt their performance. As a concrete example, F-detection was slower
if G was primed, compared to F-detection when the prime was simply the
neutral warning signal (“+”). In broader terms, it seems that priming the
“wrong” detector takes something away from the other detectors, and so
participants are worse off when they’re misled than when they receive no
prime at all.
What produces this cost? As an analogy, let’s say that you have just $50 to
spend on groceries. You can spend more on ice cream if you wish, but if you
do, you’ll have less to spend on other foods. Any increase in the ice cream
allotment, in other words, must be covered by a decrease somewhere else.
This trade-off arises, though, only because of the limited budget. If you had
unlimited funds, you could spend more on ice cream and still have enough
money for everything else.
Expectation-based priming shows the same pattern. If the Q-detector is
primed, this takes something away from the other detectors. Getting prepared for one target seems to make people less prepared for other targets. But
we just said that this sort of pattern implies a limited “budget.” If an unlimi­
ted supply of activation were available, you could prime the Q-detector and
leave the other detectors just as they were. And that is the point: Expectationbased priming, by virtue of revealing costs when misled, reveals the presence
of a limited-capacity system.
We can now put the pieces together. Ultimately, we need to explain the
facts of selective attention, including the fact that while listening to one
message you hear little content from other messages. To explain this, we’ve
proposed that perceiving involves some work, and this work requires some
limited mental resources — some process or capacity needed for performance,
but in limited supply. That’s why you can’t listen to two messages at the same
time; doing so would require more resources than you have. And now, finally,
we’re seeing evidence for those limited resources: The Posner and Snyder
research (and many other results) reveals the workings of a limited-capacity
system, just as our hypothesis demands.
Spatial Attention
The Posner and Snyder study shows that expectations about an upcoming
stimulus can influence the processing of that stimulus. But what exactly is the
nature of these expectations? How precise or vague are they?
As one way of framing this issue, imagine that participants in a study are
told, “The next stimulus will be a T.” In this case, they know exactly what to
164 •
C H A P T E R F I V E Paying Attention
get ready for. But now imagine that participants are told, “The next stimulus
will be a letter” or “The next stimulus will be on the left side of the screen.”
Will these cues allow participants to prepare themselves?
These issues have been examined in studies of spatial attention — that is,
the mechanism through which someone focuses on a particular position in
space. In one early study, Posner, Snyder, and Davidson (1980) required their
participants simply to detect letter presentations; the task was just to press
a button as soon as a letter appeared. Participants kept their eyes pointed at
a central fixation mark, and letters could appear either to the left or to the
right of this mark.
For some trials, a neutral warning signal was presented, so that participants knew a trial was about to start but had no information about stimulus
location. For other trials, an arrow was used as the warning signal. Sometimes the arrow pointed left, sometimes right; and the arrow was generally
an accurate predictor of the location of the stimulus-to-come. If the arrow
pointed right, the stimulus would be on the right side of the computer screen.
(In the terms we used earlier, this was a high-validity cue.) On 20% of the
trials, however, the arrow misled participants about location.
The results show a familiar pattern (Posner et al., 1980). With high-validity
priming, the data show a benefit from cues that correctly signal where the
upcoming target will appear. The differences between conditions aren’t large,
but keep the task in mind: All participants had to do was detect the input.
Even with the simplest of tasks, it pays to be prepared (see Figure 5.9).
What about the trials in which participants were misled? RTs in this condition were about 12% slower than those in the neutral condition. Once
again, therefore, we’re seeing evidence of a limited-capacity system. In order
to devote more attention to (say) the left position, you have to devote less
310
FIGURE 5.9 SPATIAL
ATTENTION
Mean response time (ms)
300
290
280
270
260
250
240
230
220
Expected location
(No expectation/
neutral cue)
Target appears at . . .
Unexpected location
In the Posner et al. (1980) study,
participants simply had to press
a button as soon as they saw the
target. If the target appeared in
the expected location, participants
detected it a bit more quickly. If,
however, participants were misled
about the target’s position (so that
the target appeared in an unexpected location), their responses
were slower than when the participants had no expectations at all.
Spatial Attention
•
165
attention to the right. If the stimulus then shows up on the right, you’re less
prepared for it — which is the cost of being misled.
Attention as a Spotlight
Studies of spatial attention suggest that visual attention can be compared to
a spotlight beam that can “shine” anywhere in the visual field. The “beam”
marks the region of space for which you are prepared, so inputs within the
beam are processed more efficiently. The beam can be wide or narrowly
focused (see Figure 5.10) and can be moved about at will as you explore
(i.e., attend to) various aspects of the visual field.
FIGURE 5.10
ADJUSTING THE “BEAM” OF ATTENTION
Charles Allan Gilbert’s painting All Is Vanity can be perceived either as a
woman at her dressing table or as a human skull. As you shift from one of
these perceptions to the other, you need to adjust the spotlight beam of
attention — to a narrow beam to see details (e.g., to see the woman) or to a
wider beam to see the whole scene (e.g., to see the skull).
166 •
C H A P T E R F I V E Paying Attention
Let’s emphasize, though, that the spotlight idea refers to movements of
attention, not movements of the eyes. Of course, eye movements do play
an important role in your selection of information from the world: If you
want to learn more about something, you generally look at it. (For more on
how you move your eyes to explore a scene, see Henderson, 2013; Moore
& Zirnsak, 2017.) Even so, movements of the eyes can be separated from
movements of attention, and it’s attention, not the eyes, that’s moving around
in the Posner et al. (1980) study. We know this because of the timing of the
effects. Eye movements are surprisingly slow, requiring 180 to 200 ms. But
the benefits of primes can be detected within the first 150 ms after the priming stimulus is presented. Therefore, the benefits of attention occur prior to
any eye movement, so they cannot be a consequence of eye movements.
But what does it mean to “move attention”? The spotlight beam is just a
metaphor, so we need to ask what’s really going on in the brain to produce
these effects. The answer involves a network of sites in the frontal cortex and
the parietal cortex. According to one proposal (Posner & Rothbart, 2007;
see Figure 5.11), one cluster of sites (the orienting system) is needed to disengage attention from one target, shift attention to a new target, and then
FIGURE 5.11
ANY BRAIN SITES ARE CRUCIAL
M
FOR ATTENTION
Frontal eye field
Anterior
cingulate gyrus
Frontal area
Prefrontal area
Superior parietal lobe
Posterior area
Temporoparietal
junction
Thalamus
Alerting
Orienting
Executive
Pulvinar
Superior
colliculus
Many brain sites are important for controlling attention. Some sites play a
pivotal role in alerting the brain, so that it is ready for an upcoming event.
Other sites play a key role in orienting attention, so that you’re focused on
this position or that, on one target or another. Still other sites are crucial
for controlling the brain’s executive function — a function we’ll discuss later in
the chapter.
(after posner & rothbart, 2007)
Spatial Attention
•
167
engage attention on the new target. A second set of sites (the alerting system)
is responsible for maintaining an alert state in the brain. A third set of sites
(the executive system) controls voluntary actions.
These points echo a theme we first met in Chapter 2. There, we argued
that cognitive capacities depend on the coordinated activity of multiple brain
regions, with each region providing a specialized process necessary for the
overall achievement. As a result, a problem in any of these regions can disrupt
the overall capacity, and if there are problems in several regions, the disruption can be substantial.
As an illustration of this interplay between brain sites and symptoms,
consider a disorder we mentioned earlier — ADHD. (We’ll have more to say
about ADHD later in the chapter.) Table 5.2 summarizes one proposal about
this disorder. Symptoms of ADHD are listed in the left column; the right
column identifies brain areas that may be the main source of each symptom.
This proposal is not the only way to think about ADHD, but it illustrates
the complex, many-part relationship between overall function (in this case,
the ability to pay attention) and brain anatomy. (For more on ADHD, see
Barkley, Murphy, & Fischer, 2008; Brown, 2005; Seli, Smallwood, Cheyne,
& Smilek, 2015; Zillmer, Spiers, & Culbertson, 2008.)
In addition, the sites listed in Table 5.2 can be understood roughly as
forming the “control system” for attention. Entirely different sites (including
TABLE 5.2
ATTENTION-DEFICIT/HYPERACTIVITY DISORDER
SYMPTOMS, COGNITIVE PROCESSES, AND
NEURAL NETWORKS
Symptom Domains and Cognitive Processes
Relevant Brain Site
Problems in the “Alerting” system
Has difficulty sustaining attention
Right frontal cortex
Fails to finish
Right posterior parietal
Avoids sustained efforts
Locus ceruleus
Problems in the “Orienting” system
Is distracted by stimuli
Bilateral parietal
Does not appear to listen
Superior colliculus
Fails to pay close attention
Thalamus
Problems in the “Executive” system
Blurts out answers
Anterior cingulate
Interrupts or intrudes
Left lateral frontal
Cannot wait
Basal ganglia
Earlier in the chapter, we mentioned the disorder known as ADHD. The table summarizes one influential proposal about this disorder, linking the symptoms of ADHD to
the three broad processes (alerting, orienting, and executive) described in the text,
and then linking these processes to relevant brain areas (Swanson et al., 2000; for a
somewhat different proposal, though, see Barkley, Murphy, & Fischer, 2008).
168 •
C H A P T E R F I V E Paying Attention
FIGURE 5.12
S ELECTIVE ATTENTION ACTIVATES THE VISUAL CORTEX
Attend left > Attend right
Left
hemisphere
Attend right > Attend left
Right
hemisphere
A
B
The brain sites that control attention are separate from the brain sites that do the actual analysis of the input. Thus,
the intention to attend to, say, stimuli on the left is implemented through the many brain sites shown in Figure 5.11.
However, these sites collectively activate a different set of sites — in the visual cortex — to promote the actual processing of the incoming stimulus. Shown here are activity levels in one participant (measured through fMRI scans)
overlaid on a structural image of the brain (obtained through MRI scans). Keep in mind that because of the brain’s
contralateral organization, the intention to pay attention to the left side of space requires activation in the right
hemisphere (Panel A); the intention to pay attention to the right requires activation in the left (Panel B).
the visual areas in the occipital cortex) do the actual analysis of the incoming
information (see Figure 5.12). In other words, neural connections from the
areas listed in the table carry signals to the brain regions that do the work
of analyzing the input. These control signals can amplify (or, in some cases,
inhibit) the activity in these other areas and, in this way, they can promote
the processing of inputs you’re interested in, and undermine the processing of
distractors. (See Corbetta & Shulman, 2002; Hampshire, Duncan, & Owen,
2007; Hon, Epstein, Owen, & Duncan, 2006; Hung, Driver, & Walsh, 2005;
Miller & Cohen, 2001.)
Thus, there is no spotlight beam. Instead, certain neural mechanisms
enable you to adjust your sensitivity to certain inputs. This is, of course,
entirely in line with the proposal we’re developing — namely, that a large
part of “paying attention” involves priming. For stimuli you don’t care
about, you don’t bother to prime yourself, and so those stimuli fall on
unprepared (and unresponsive) detectors. For stimuli you do care about,
you do your best to anticipate the input, and you use these anticipations
to prime the relevant processing channel. This increases your sensitivity
to the desired input, which is just what you want. (For further discussion
of the “spotlight” idea, see Cave, 2013; Rensink, 2012; Wright & Ward,
2008; but also Awh & Pashler, 2000; Morawetz, Holz, Baudewig, Treue,
& Dechent, 2007.)
Spatial Attention
•
169
Where Do We “Shine” the “Beam”?
So far, we’ve been discussing how people pay attention, but we can also ask
what people pay attention to. Where do people “shine” the “spotlight beam”?
The answer has several parts. As a start, you pay attention to elements of
the input that are visually prominent (Parkhurst, Law, & Niebur, 2002) and
also to elements that you think are interesting or important. Decisions about
what’s important, though, depend on the context. For example, Figure 5.13
shows classic data recording a viewer’s eye movements while inspecting a
picture (Yarbus, 1967). The target picture is shown in the top left. Each of the
other panels shows a three-minute recording of the viewer’s eye movements;
FIGURE 5.13
EYE MOVEMENTS AND VISION
1
2
Free examination.
3
Give the ages of the people.
Estimate material circumstances
of the family.
4
5
Remember the clothes
worn by the people.
Surmise what the family had
been doing before the arrival
of the unexpected visitor.
6
Remember positions of people and
objects in the room.
7
Estimate how long the visitor had
been away from the family.
Participants were shown the picture in the top left. Each of the other panels shows a three-minute recording
of one viewer’s eye movements while inspecting the picture. The labels for each panel summarize the viewer’s
goal while looking at the picture. Plainly, the pattern of the movements depended on what the viewer was trying
to learn.
170 •
C H A P T E R F I V E Paying Attention
plainly, the pattern of movements depended on what the viewer was trying
to learn about the picture.
In addition, your beliefs about the scene play an important role. You’re
unlikely to focus, for example, on elements of a scene that are entirely predictable, because you’ll gain no information from inspecting things that are
already obvious (Brewer & Treyens, 1981; Friedman, 1979; Võ & Henderson,
2009). But you’re also unlikely to focus on aspects of the scene that are
totally unexpected. If, for example, you’re walking through a forest, you
won’t be on the lookout for a stapler sitting on the ground, and so you may
fail to notice the stapler (unless it’s a bright color, or directly in your path, or
some such). This point provides part of the basis for inattentional blindness
(pp. 155–156) and also leads to a pattern called the “ultra-rare item effect”
(Mitroff & Biggs, 2014). The term refers to a pattern in which rare items
are often overlooked; as the authors of one paper put it, “If you don’t find it
often, you often don’t find it” (Evans, Birdwell, & Wolfe, 2013; for a troubling consequence of this pattern, see Figure 5.14).
As another complication, people differ in what they pay attention to
(e.g., Castelhano & Henderson, 2008), although some differences aren’t surprising. For example, in looking at a scene, women are more likely than men
to focus on how the people within the scene are dressed; men are more likely
to focus on what the people look like (including their body shapes; Powers,
Andriks, & Loftus, 1979).
Perhaps more surprising are differences from one culture to the next
in how people pay attention. The underlying idea here is that people in
the West (the United States, Canada, most of Europe) live in “individualistic” cultures that emphasize the achievements and qualities of the single
person; therefore, in thinking about the world, Westerners are likely to
focus on individual people, individual objects, and their attributes. In contrast, people in East Asia have traditionally lived in “collectivist” cultures
that emphasize the ways in which all people are linked to, and shaped
by, the people around them. East Asians are therefore encouraged to
think more holistically, with a focus on the context and how people and
FIGURE 5.14
IF YOU DON’T FIND IT OFTEN . . .
Data both in the laboratory and in real-world settings tell us that
people often overlook targets if the targets happen to be quite
rare. As one group of authors put it, “If you don’t find it often, you
often don’t find it.” This pattern has troubling implications for the
security inspections routinely conducted at airports: The inspectors
will see troubling items only rarely and, as a result, are likely to overlook those troubling items.
Spatial Attention
•
171
objects are related to one another. (See Nisbett, 2003; Nisbett, Peng, Choi, &
Norenzayan, 2001; also Tardif et al., 2017. For a broad review, see
Heine, 2015.)
This linkage between culture and cognition isn’t rigid, and so people in
any culture can stray from these patterns. Even so, researchers have docu­
mented many manifestations of these differences from one culture to the
next — including differences in how people pay attention. In one study,
researchers tracked participants’ eye movements while the participants were
watching animated scenes on a computer screen (Masuda et al., 2008). In the
initial second of viewing, there was little difference between the eye movements of American participants and Japanese participants: Both groups spent
90% of the time looking at the target person, located centrally in the display.
But in the second and third seconds of viewing, the groups differed. The
Americans continued to spend 90% of their time looking directly at the central figure (and so spent only 10% of their time looking at the faces of people
visible in the scene’s background); Japanese participants, in contrast, spent
between 20% and 30% of their time looking at the faces in the background.
This difference in viewing time was reflected in the participants’ judgments
about the scene. When asked to make a judgment about the target person’s
emotional state, American participants weren’t influenced by the emotional
expressions of the people standing in the background, but Japanese participants were.
In another study, American and East Asian students were tested in a
change-blindness procedure. They viewed one picture, then a second, then
the first again, and this alternation continued until the participants spotted the difference between the two pictures. For some pairs of pictures, the
difference involved an attribute of a central object in the scene (e.g., the
color of a truck changed from one picture to the next). For these pictures,
Americans and East Asians performed equivalently, needing a bit more than
9 seconds to spot the change. For other pairs of pictures, the difference
involved the context (e.g., a pattern of the clouds in the background of
the scene). For these pictures, the East Asians were notably faster than the
Americans in detecting the change. (See Masuda & Nisbett, 2006. For other
data, exploring when Westerners have a performance advantage and when
East Asians have the advantage, see Amer, Ngo, & Hasher, 2016; Boduroglu,
Shah, & Nisbett, 2010.)
Finally, let’s acknowledge that sometimes you choose what to pay
attention to — a pattern called endogenous control of attention. But sometimes an element of the scene “seizes” your attention whether you like it or
not, and this pattern is called exogenous control of attention. Exogenous
control is of intense interest to theorists, and it’s also important for
pragmatic reasons. For example, people who design ambulance sirens or
warning signals in an airplane cockpit want to make sure these stimuli
cannot be ignored. In the same way, advertisers do all they can to ensure
that their product name or logo will grab your attention even if you’re
172 •
C H A P T E R F I V E Paying Attention
FIGURE 5.15
EXOGENOUS CONTROL OF ATTENTION
Public health officials would like the health warning shown here to seize your attention, so that you can’t
overlook it. The tobacco industry, however, might have a different preference.
intensely focused on something else (e.g., the competitor’s product; also
see Figure 5.15).
Attending to Objects or Attending to Positions
A related question is also concerned with the “target” of the attention “spotlight.” To understand the issue, think about how an actual spotlight works. If
a spotlight shines on a donut, then part of the beam will fall on the donut’s
hole and will illuminate part of the plate underneath the donut. Similarly, if
the beam isn’t aimed quite accurately, it may also illuminate the plate just to
the left of the donut. The region illuminated by the beam, in other words, is
defined purely in spatial terms: a circle of light at a particular position. That
circle may or may not line up with the boundaries of the object you’re shining
the beam on.
Is this how attention works — so that you pay attention to whatever falls
in a certain region of space? If this is the case, you might at times end up
paying attention to part of this object, part of that. An alternative is that you
pay attention to objects rather than to positions in space. To continue the
example, the target of your attention might be the donut itself rather than
Spatial Attention
•
173
FIGURE 5.16
UNILATERAL NEGLECT SYNDROME
A patient with damage to the right parietal cortex was
asked to draw a typical clock face. In his drawing, the patient
seemed unaware of the left side, but he still recalled that all
12 numbers had to be displayed. The drawing shows how he
resolved this dilemma.
its location. In that case, the plate just to the left and the bit of plate visible
through the donut’s hole might be close to your focus, but they aren’t part of
the attended object and so aren’t attended.
Which is the correct view of attention? Do you pay attention to regions
in space, no matter what objects (or parts of objects) fall in that region? Or
do you pay attention to objects? It turns out that each view captures part of
the truth.
One line of evidence comes from the study of people we mentioned at
the chapter’s start – people who suffer from unilateral neglect syndrome (see
Figure 5.16). Taken at face value, the symptoms shown by these patients seem
to support a space-based account of attention: The afflicted patient seems
insensitive to all objects within a region that’s defined spatially — namely,
everything to the left of his or her current focus. If an object falls half within
the region and half outside of it, then the spatial boundary is what matters, not the object’s boundaries. This is clear, for example, in how these
patients read words (likely to read “BOTHER” as “HER” or “CARROT” as
“ROT”) — responding only to the word’s right half, apparently oblivious to
the word’s overall boundaries.
Other evidence, however, demands further theory. In one study, patients
with neglect syndrome had to respond to targets that appeared within a
barbell-shaped frame (see Figure 5.17). Not surprisingly, they were much
more sensitive to the targets appearing within the red circle (on the right) and
missed many of the targets appearing in the blue circle (on the left); this result
confirms the patients’ diagnosis. What’s crucial, though, is what happened
next. While the patients watched, the barbell frame was slowly spun around,
174 •
C H A P T E R F I V E Paying Attention
FIGURE 5.17
SPACE-BASED OR OBJECT-BASED ATTENTION
Patient initially sees:
As the patient watches:
Patient now sees:
Patients with unilateral neglect syndrome were much more sensitive to
targets appearing within the red circle (on the right) and missed many of
the targets appearing within the blue circle (on the left); this observation
confirms their clinical diagnosis. Then, as the patients watched, the barbellshaped frame rotated, so that now the red circle was on the left and the blue
circle was on the right. After this rotation, participants were still more sensitive to targets in the red circle (now on the left), apparently focusing on this
attended object even though it had moved into their “neglected” side.
so that the red circle, previously on the right, was now on the left and the blue
circle, previously on the left, was now on the right.
If the patients consistently neglect a region of space, they should now be
more sensitive to the (right-side) blue circle. But here’s a different possibility:
Perhaps these patients have a powerful bias to attend to the right side, and so
initially they attend to the red circle. Once they have “locked in” to this circle,
however, it’s the object, not the position in space, that defines their focus of
attention. According to this view, if the barbell rotates, they will continue
attending to the red circle (this is, after all, the focus of their attention), even
though it now appears on their “neglected” side. This prediction turns out to
be correct: When the barbell rotates, the patients’ focus of attention seems to
rotate with it (Behrmann & Tipper, 1999).
To describe these patients, therefore, we need a two-part account. First, the
symptoms of neglect syndrome plainly reveal a spatially defined bias: These
Spatial Attention
•
175
patients neglect half of space. But, second, once attention is directed toward
a target, it’s the target itself that defines the focus of attention; if the target
moves, the focus moves with it. In this way, the focus of attention is objectbased, not space-based. (For more on these issues, see Chen & Cave, 2006;
Logie & Della Salla, 2005; Richard, Lee, & Vecera, 2008.)
And it’s not just people with brain damage who show this complex pattern. People with intact brains also show a mix of space-based and objectbased attention. We’ve already seen evidence for the spatial base: The Posner
et al. (1980) study and many results like it show that participants can focus
on a particular region of space in preparation for a stimulus. In this situation,
the stimulus has not yet appeared; there is no object to focus on. Therefore,
the attention must be spatially defined.
In other cases, though, attention is heavily influenced by object boundaries. For example, in some studies, participants have been shown displays with
visually superimposed stimuli (e.g., Becklen & Cervone, 1983; Neisser &
Becklen, 1975). Participants can usually pay attention to one of these stimuli and ignore the other. This selection cannot be space-based (because both
stimuli are in the same place) and so must be object-based.
This two-part account is also supported by neuroscience evidence. Various studies have examined the pattern of brain activation when participants are attending to a particular position in space, and the pattern of
activation when participants are attending to a particular object. These
data suggest that the tasks involve different brain circuits — with one set
of circuits (the dorsal attention system), near the top of the head, being
primarily concerned with spatial attention, and a different set of circuits
(the ventral attention system) being crucial for nonspatial tasks (Cave,
2013; Cohen, 2012; Corbetta & Shulman, 2011). Once again, therefore,
our description of attention needs to include a mix of object-based and
space-based mechanisms.
Perceiving and the Limits on Cognitive Capacity:
An Interim Summary
Let’s pause to review. In some circumstances, you seem to inhibit the processing of unwanted inputs. This inhibitory process is quite specific (the inhibition
blocks the processing of a particular well-defined input) and certainly benefits
from practice.
More broadly, though, various mechanisms facilitate the processing of
desired inputs. The key here is priming. You’re primed for some stimuli
because you’ve encountered them often in the past, with the result that you’re
more likely to process (and therefore more likely to notice) these stimuli if
you encounter them again. In other cases, the priming depends on your ability to anticipate what the upcoming stimulus will be. If you can predict what
the stimulus will likely be, you can prime the relevant processing pathway
so that you’re ready for the stimulus when it arrives. The priming will make
you more responsive to the anticipated input if it does arrive, and this gives
176 •
C H A P T E R F I V E Paying Attention
the anticipated input an advantage relative to other inputs. That advantage
is what you want — so that you end up perceiving the desired input (the one
you’ve prepared yourself for) but don’t perceive the inputs you’re hoping to
ignore (because they fall on unprimed detectors).
The argument, then, is that your ability to pay attention depends to a large
extent on your ability to anticipate the upcoming stimulus. This anticipation,
in turn, depends on many factors. You’ll have a much easier time anticipating
(and so an easier time paying attention to) materials that you understand as
opposed to materials that you don’t understand. Likewise, when a stimulus
sequence is just beginning, you have little basis for anticipation, so your only
option may be to focus on the position in space that holds the sequence. Once
the sequence begins, though, you get a sense of how it’s progressing, and this
lets you sharpen your anticipations (shifting to object-based attention, rather
than space-based) — which, again, makes you more sensitive to the attended
input and more resistant to distractors.
Putting all these points together, perhaps it’s best not to think of the term
“attention” as referring to a particular process or a particular mechanism.
Instead, it’s better to think of attention (as one research group put it) as an
effect rather than as a cause (Krauzlis, Bollimunta, Arcizet, & Wang, 2014).
In other words, the term “attention” doesn’t refer to some mechanism in the
brain that produces a certain outcome. It’s better to think of attention as itself
an outcome — a byproduct of many other mechanisms.
As a related perspective, it may be helpful to think of paying attention, not as a process or mechanism, but as an achievement — something
that you’re able to do. Like many other achievements (e.g., doing well in
school, staying healthy), paying attention involves many elements, and the
exact set of elements needed will vary from one occasion to the next. In all
cases, though, multiple steps are needed to ensure that you end up being
aware of the stimuli you’re interested in, and not getting pulled off track by
irrelevant inputs.
TEST YOURSELF
6.In what ways does the
notion of a spotlight
beam accurately
reflect how spatial
attention functions?
7.In what ways does the
notion of a spotlight
beam differ from the
way spatial attention
functions?
8.When you first start
paying attention to an
input, your attention
seems to be spacebased. Once you’ve
learned a bit about
the input, though,
your attention seems
to be object-based.
How does this pattern
fit with the idea that
you pay attention by
anticipating the input?
Divided Attention
So far in this chapter, we’ve emphasized situations in which you’re trying to
focus on a single input. If other tasks and other stimuli were on the scene,
they were mere distractors. Sometimes, though, your goal is different: You
want to “multitask” — that is, deal with multiple tasks, or multiple inputs, all
at the same time. What can we say about this sort of situation — a situation
involving divided attention?
Sometimes divided attention is easy. Almost anyone can walk and sing
simultaneously; many people like to knit while they’re holding a conversation or listening to a lecture. It’s much harder, though, to do calculus homework while listening to a lecture; and trying to get your assigned reading
done while watching TV is surely a bad bet. What lies behind this pattern?
Why are some combinations difficult while others are easy?
Divided Attention
•
177
Our first step toward answering these questions is already in view. We’ve
proposed that perceiving requires resources that are in limited supply; the
same is presumably true for other tasks — remembering, reasoning, problem
solving. They, too, require resources, and without these resources the processes cannot go forward. What are the resources? The answer includes a
mix of things: certain mechanisms that do specific jobs, certain types of
memory that hold on to information while you’re working on it, energy supplies to keep the mental machinery going, and more. No matter what the
resources are, though, a task will be possible only if you have the needed
resources — just as a dressmaker can produce a dress only if he has the raw
materials, the tools, the time needed, the energy to run the sewing machine,
and so on.
All of this leads to a proposal: You can perform concurrent tasks only if
you have the resources needed for both. If the two tasks, when combined,
require more resources than you’ve got, then divided attention will fail.
The Specificity of Resources
CAESAR THE
MULTITASKER
Some writers complain about
the hectic pace of life today
and view it as a sad fact
about the pressured reality of the modern world. But
were things different in earlier times? More than 2,000
years ago, Julius Caesar was
praised for his ability to multi­
task. (The term is new, but
the capacity is not.) According to the Roman historian
Suetonius,
Caesar
could
write, dictate letters, and read
at the same time. Even on the
most important subjects, he
could dictate four letters at
once — and if he had nothing
else to do, as many as seven
letters at once. From a modern perspective, though, we
can ask: Is any of this plausible? Perhaps it is — some
people do seem especially
skilled at multitasking (Just &
Buchweitz, 2017; Redick et al.,
2016), and maybe Caesar was
one of those special people!
178 •
Imagine that you’re hoping to read a novel while listening to an academic
lecture. These tasks are different, but both involve the use of language, and
so it seems likely that these tasks will have overlapping resource requirements. As a result, if you try to do the tasks at the same time, they’re likely
to compete for resources — and therefore this sort of multitasking will
be difficult.
Now, think about two very different tasks, such as knitting and listening
to a lecture. These tasks are unlikely to interfere with each other. Even if all
your language-related resources are in use for the lecture, this won’t matter
for knitting, because it’s not a language-based task.
More broadly, the prediction here is that divided attention will be easier
if the simultaneous tasks are very different from each other, because different
tasks are likely to have distinct resource requirements. Resources consumed
by Task 1 won’t be needed for Task 2, so it doesn’t matter for Task 2 that
these resources are tied up in another endeavor.
Is this the pattern found in the research data? In an early study by Allport,
Antonis, and Reynolds (1972), participants heard a list of words presented through headphones into one ear, and their task was to shadow (i.e.,
repeat back) these words. At the same time, they were also presented with
a second list. No immediate response was required to the second list, but
later on, memory was tested for these items. In one condition, the second
list (the memory items) consisted of words presented into the other ear, so
the participants were hearing (and shadowing) a list of words in one ear
while simultaneously hearing the memory list in the other. In another condition, the memory items were presented visually. That is, while the participants were shadowing one list of words, they were also seeing a different list
of words on a screen before them. Finally, in a third condition, the memory
items consisted of pictures, also presented on a screen.
C H A P T E R F I V E Paying Attention
These three conditions had the same requirements — shadowing one list
while memorizing another. But the first condition (hear words + hear words)
involved very similar tasks; the second condition (hear words + see words)
involved less similar tasks; the third condition (hear words + see pictures),
even less similar tasks. On the logic we’ve discussed, we should expect the
most interference in the first condition and the least interference in the third.
And that is what the data showed (see Figure 5.18).
The Generality of Resources
Similarity among tasks, however, is not the whole story. If it were, then we’d
observe less and less interference as we consider tasks further and further apart.
Eventually, we’d find tasks so different from each other that we’d observe no
interference at all between them. But that’s not the pattern of the evidence.
Consider the common practice of talking on a cell phone while driving.
When you’re on the phone, the main stimulus information comes into your
ear, and your primary response is by talking. In driving, the main stimulation comes into your eyes, and your primary response involves control of
your hands on the steering wheel and your feet on the pedals. For the phone
conversation, you’re relying on language skills. For driving, you need spatial
skills. Overall, it looks like there’s little overlap in the specific demands of
these two tasks, and so little chance that the tasks will compete for resources.
Data show, however, that driving and cell-phone use do interfere with
each other. This is reflected, for example, in the fact that phone use has
been implicated in many automobile accidents (Lamble, Kauranen, Laakso,
& Summala, 1999). Even with a hands-free phone, drivers engaged in cellphone conversations are more likely to be involved in accidents, more likely
to overlook traffic signals, and slower to hit the brakes when they need to.
Errors in recognition
50
40
30
FIGURE 5.18 DIVIDED ATTENTION AMONG
DISTINCT TASKS
20
10
0
Words
Words Pictures
heard
seen
Type of remembered items
Participants perform poorly if they are trying to shadow one
list of words while hearing other words. They do somewhat
better if shadowing while seeing other words. They do better still
if shadowing while seeing pictures. In general, the greater the
difference between two tasks, the easier it will be to combine
the tasks.
(after allport, antonis, & reynolds, 1972)
Divided Attention
•
179
(See Kunar, Carter, Cohen, & Horowitz, 2008; Levy & Pashler, 2008;
Sanbonmatsu, Strayer, Biondi, Behrends, & Moore, 2016; Stothart, Mitchum,
& Yehnert, 2015; Strayer & Drews, 2007; Strayer, Drews, & Johnston, 2003.
For some encouraging data, though, on why phone-related accidents don’t
occur even more often, see Garrison & Williams, 2013; Medeiros-Ward,
Watson, & Strayer, 2015.)
As a practical matter, therefore, talking on the phone while driving is a
bad idea. In fact, some people estimate that the danger caused by driving
while on the phone is comparable to (and perhaps greater than) the risk of
driving while drunk. But, on the theoretical side, notice that the interference
observed between driving and talking is interference between two hugely
distinctive activities — a point that provides important information about
the nature of the resource competition involved, and therefore the nature of
mental resources.
Before moving on, we should mention that the data pattern is different
if the driver is talking to a passenger in the car rather than using the phone.
Conversations with passengers seem to cause little interference with driving (Drews, Pasupathi, & Strayer, 2008; see Figure 5.19), and the reason is
simple. If the traffic becomes complicated or the driver has to perform some
FIGURE 5.19
CELL PHONE USE AND DRIVING
.04
.02
A
No cell
phone
500
400
With cell
phone
Success in simple highway
navigation (percentage)
.06
0
100
600
Reaction time (ms)
Fraction of lights missed
.08
B
No cell
phone
90
80
70
60
50
40
30
20
10
0
With cell
phone
C
Drive while
talking with
a passenger
Drive while
conversing
via cell phone
Many studies show that driving performance is impaired when the driver is on the phone (whether hand-held or
hands-free). While on the phone, drivers are more likely to miss a red light (Panel A) and are slower in responding
to a red light (Panel B). Disruption is not observed, however, if the driver is conversing with a passenger rather
than on the phone (Panel C). That’s because the passenger is likely to adjust her conversation to accommodate
changes in driving — such as not speaking while the driver is navigating an obstruction.
(after strayer & johnston, 2001)
180 •
C H A P T E R F I V E Paying Attention
tricky maneuver, the passenger can see this — either by looking out of the
car’s window or by noticing the driver’s tension and focus. In these cases,
passengers helpfully slow down their side of the conversation, which takes
the load off of the driver, enabling the driver to focus on the road (Gaspar
et al., 2014; Hyman, Boss, Wise, McKenzie, & Caggiano, 2010; Nasar, Hecht,
& Wener, 2008).
Executive Control
The evidence is clear, then, that tasks as different as driving and talking compete with each other for some mental resource. But what is this resource,
which is apparently needed for both verbal tasks and spatial ones, tasks with
visual inputs and tasks with auditory inputs?
Evidence suggests that multiple resources may be relevant. Some authors
describe resources that serve (roughly) as an energy supply, drawn on by
all tasks (Eysenck, 1982; Kahneman, 1973; Lavie, 2001, 2005; Lavie, Lin,
Zokaei, & Thoma, 2009; MacDonald & Lavie, 2008; Murphy, Groeger, &
Greene, 2016). According to this perspective, tasks vary in the “load” they
put on you, and the greater the load, the greater the interference with other
tasks. In one study, drivers were asked to estimate whether their vehicle
would fit between two parked cars (Murphy & Greene, 2016). When the
judgment was difficult, participants were less likely to notice an unexpected
pedestrian at the side of the road. In other words, higher perceptual load
(from the driving task) increased inattentional blindness. (Also see Murphy
& Greene, 2017.)
Other authors describe mental resources as “mental tools” rather than
as some sort of mental “energy supply.” (See Allport, 1989; Baddeley, 1986;
Bourke & Duncan, 2005; Dehaene, Sergent, & Changeux, 2003; JohnsonLaird, 1988; Just, Carpenter, & Hemphill, 1996; Norman & Shallice, 1986;
Ruthruff, Johnston, & Remington, 2009; Vergaujwe, Barrouillet, & Camos,
2010. For an alternative perspective on these resources, see Franconeri,
2013; Franconeri, Alvarez, & Cavanagh, 2013.) One of these tools is
especially important, and it involves the mind’s executive control. This term
refers to the mechanisms that allow you to control your own thoughts, and
these mechanisms have multiple functions. Executive control helps keep your
current goals in mind, so that these goals (and not habit) will guide your
actions. The executive also ensures that your mental steps are organized into
the right sequence — one that will move you toward your goals. And if your
current operations aren’t moving you toward your goal, executive control
allows you to shift plans, or change strategy. (For discussion of how the
executive operates and how the brain tissue enables executive function, see
Brown, Reynolds, & Braver, 2007; Duncan et al., 2008; Gilbert & Shallice,
2002; Kane, Conway, Hambrick, & Engle, 2007; Miller & Cohen, 2001;
Miyake & Friedman, 2012; Shipstead, Harrison, & Engle, 2015; Stuss &
Alexander, 2007; Unsworth & Engle, 2007; Vandierendonck, Liefooghe, &
Verbruggen, 2010.)
CELL-PHONE DANGERS
FOR PEDESTRIANS
It’s not just driving that’s
disrupted by cell-phone use.
Compared to pedestrians
who aren’t using a phone,
pedestrians engaged in phone
conversations tend to walk
more slowly and more
erratically, and are less likely
to check traffic before they
cross a street. They’re also
less likely to notice things
along their path. In one
study, researchers observed
pedestrians walking across a
public square (Hyman, Boss,
Wise, McKenzie, & Caggiano,
2010). If the pedestrian was
walking with a friend (and
so engaged in a “live” conversation), there was a 71%
chance the pedestrian would
notice the unicycling clown
just off the pedestrian’s path.
But if the pedestrian was on
the phone (i.e., engaging in a
telephonic conversation), the
person had only a 25% chance
of detecting the clown.
Divided Attention
•
181
Executive control can only handle one task at a time, and this point
obviously puts limits on your ability to multitask — that is, to divide your
attention. But executive control is also important when you’re trying to do
just a single task. Evidence comes from studies of people who have suffered
damage to the prefrontal cortex (PFC), a brain area right behind the eyes
that seems crucial for executive control. People with this damage (including
Phineas Gage, whom we met in Chapter 2) can lead relatively normal lives,
because in their day-to-day behavior they can often rely on habit or can simply respond to prominent cues in their environment. With appropriate tests,
though, we can reveal the disruption that results from frontal lobe damage.
In one commonly used task, patients with frontal lesions are asked to sort a
deck of cards into two piles. At the start, the patients have to sort the cards
according to color; later, they need to switch strategies and sort according
to the shapes shown on the cards. The patients have enormous difficulty in
making this shift and continue to sort by color, even though the experimenter
tells them again and again that they’re placing the cards on the wrong piles
(Goldman-Rakic, 1998). This is referred to as a perseveration error, a tendency to produce the same response over and over even when it’s plain that
the task requires a change in the response.
These patients also show a pattern of goal neglect — failing to organize
their behavior in a way that moves them toward their goals. For example,
when one patient was asked to copy Figure 5.20A, the patient produced
the drawing shown in Figure 5.20B. The copy preserves features of the
FIGURE 5.20
A
GOAL NEGLECT
B
C
Patients who had suffered damage to the prefrontal cortex were asked to copy the drawing in Panel A. One
patient’s attempt is shown in Panel B; the drawing is reasonably accurate but seems to have been drawn with
no overall plan — for example, the large rectangle in the original, and the main diagonals, were created piecemeal rather than being used to organize the drawing. Another patient’s attempt is shown in Panel C; this patient
started to re-create the drawing but then got swept up in her own artistic impulses.
182 •
C H A P T E R F I V E Paying Attention
original, but close inspection reveals that the patient drew the copy with
no particular plan in mind. The large rectangle that defines the shape was
never drawn, and the diagonal lines that organize the figure were drawn in a
piecemeal fashion. Many details are correctly reproduced but weren’t drawn
in any sort of order; instead, these details were added whenever they happened to catch the patient’s attention (Kimberg, D’Esposito, & Farah, 1998).
Another patient, asked to copy the same figure, produced the drawing shown
in Figure 5.20C. This patient started to draw the figure in a normal way,
but then she got swept up in her own artistic impulses, adding stars and a
smiley face (Kimberg et al., 1998). (For more on executive control, see Aron,
2008; Courtney, Petit, Maisog, Ungerleider, & Haxby, 1998; Duncan et al.,
2008; Gilbert & Shallice, 2002; Huey, Krueger, & Grafman, 2006; Kane &
Engle, 2003; Kimberg et al., 1998; Logie & Della Salla, 2005; Ranganath &
Blumenfeld, 2005; Stuss & Levine, 2002; also Chapter 13.)
Divided Attention: An Interim Summary
Our consideration of selective attention drove us toward a several-part
account, with one mechanism serving to block out unwanted distractors and
other mechanisms promoting the processing of interesting stimuli. Now, in
our discussion of divided attention, we again need several elements in our
theory. Interference between tasks is increased if the tasks are similar to each
other, presumably because similar tasks overlap in their processing requirements and make competing demands on mental resources that are specialized
for that sort of task.
But interference can also be observed with tasks that are entirely different
from each other — such as driving and talking on a cell phone. Therefore, our
account needs to include resources that are general enough in their use that
they’re drawn on by almost any task. We’ve identified several of these general
resources: an energy supply needed for mental tasks, executive control, and
others as well. No matter what the resource, though, the key principle will
be the same: Tasks will interfere with each other if their combined demand
for a resource is greater than the amount available — that is, if the demand
exceeds the supply.
TEST YOURSELF
9.Why is it easier to
divide attention
between very different
activities (e.g., knitting while listening
to a lecture) than it
is to divide attention
between more similar
activities?
10.What is executive
control, and why does
it create limits on your
ability to divide your
attention between two
simultaneous tasks?
Practice
For a skilled driver, talking on a cell phone while driving is easy as long as
the driving is straightforward and the conversation is simple. Things fall
apart, though, the moment the conversation becomes complex or the driving becomes challenging. Engaged in deep conversation, the driver misses
a turn; while maneuvering through an intersection, the driver suddenly
stops talking.
The situation is different, though, for a novice driver. For someone who’s
just learning to drive, driving is difficult all by itself, even on a straight
road with no traffic. If we ask the novice to do anything else at the same
Practice
•
183
COGNITION
outside the lab
“I Can’t Ignore . . .”
At the start of this chapter, we refer to common
won’t be especially sensitive to the input when
experiences like this one: You’re trying to read a
it arrives.
book — perhaps an assignment for one of your
What about the conversation you’re overhear-
courses. You’re in a public place, though, and two
ing and hoping to ignore? Maybe it’s unfolding
people nearby are having a conversation. You have
according to a familiar script — for example, the
no interest in their conversation and really need
people behind you on the bus, or on the other side
to get through your reading assignment. Even so,
of the room, are discussing romance or a popular
you find yourself unable to ignore their conver-
movie. In this setting, with almost no thought you’ll
sation, so your reading doesn’t get done. What’s
easily anticipate where this distractor conversation
going on here? Why can’t you control what you’re
is going, and the anticipation will prime the rele-
paying attention to?
vant nodes in your mind, making you more sensi-
Think about the mechanisms that allow you to
tive to the input — the opposite of what you want.
pay attention. When you’re reading or listening
Part of our explanation, then, lies in ease-
to something, you do what you can to anticipate
of-anticipation. That’s why you’ll probably avoid
the upcoming input, and that anticipation lets
distraction if the material you’re trying to read is
you prime the relevant detectors so that they’ll
something you can anticipate (and so prime your-
be ready when the input arrives. As a result, the
self for) and if the irrelevant conversation involves
input falls onto “prepared” detectors and so you’re
content you can’t easily anticipate. In the extreme,
more sensitive to the input — more likely to notice
imagine that the irrelevant conversation is in some
it, more likely to process it.
foreign language that you don’t speak; here, be-
In contrast, you won’t try to anticipate inputs
you don’t care about. With no anticipation, the
cause there’s no basis for anticipation, the distraction will be minimal.
input falls on unprepared detectors — and so the
However, we need another element in our expla-
detectors are less sensitive to the input. In other
nation. Most people aren’t distracted if they try to
words, you’ve done nothing to make yourself
read while music is playing in the room or if there are
sensitive to these inputs, and so you’re relatively
traffic noises in the background. Why don’t these
insensitive to them, just as you wish.
(potential) distractors cause problems? Here, the
Now think about the situation with which
key is resource competition. Reading a book and
we began. Perhaps you’re trying to read some-
hearing a conversation both involve language, so
thing challenging. You’ll therefore have some
these activities draw on the same mental resources
difficulty anticipating how the passage will
and compete for those resources. But reading a
unfold — what words or phrases will be coming
book and hearing music (especially instrumen-
up. Therefore, you’ll have little basis for priming
tal music) draw on different mental resources, so
the soon-to-be-needed detectors, and so you
those activities don’t compete for resources.
184 •
C H A P T E R F I V E Paying Attention
time — whether it’s talking on a cell phone or even listening to the radio — we
put the driver (and other cars) at substantial risk. Why is this? Why are things
so different after practice?
Practice Diminishes Resource Demand
We’ve already said that mental tasks require resources, with the particular
resources required being dependent on the nature of the task. Let’s now add
another claim: As a task becomes more practiced, it requires fewer resources,
or perhaps it requires less frequent use of these resources.
This decrease in a task’s resource demand may be inevitable, given the
function of some resources. Consider executive control. We’ve mentioned
that this control plays little role if you can rely on habit or routine in performing a task. (That’s why Phineas Gage was able to live a mostly normal
life, despite his brain damage.) But early in practice, when a task is new, you
haven’t formed any relevant habits yet, so you have no habits to fall back on.
As a result, executive control is needed all the time. Once you’ve done the
task over and over, though, you do acquire a repertoire of suitable habits, and
so the demand for executive control decreases.
How will this matter? We’ve already said that tasks interfere with each
other if their combined resource demand is greater than the amount of resources available. Interference is less likely, therefore, if the “cognitive cost”
of a task is low. In that case, you’ll have an easier time accommodating the
task within your “resource budget.” And we’ve now added the idea that the
resource demand (the “cost”) will be lower after practice than before. Therefore, it’s no surprise that practice makes divided attention easier — enabling
the skilled driver to continue chatting with her passenger as they cruise down
the highway, even though this combination is hopelessly difficult for the
novice driver.
Automaticity
With practice in a task, then, the need for executive control is diminished, and
we’ve mentioned one benefit of this: Your control mechanisms are available
for other chores, allowing you to divide your attention in ways that would
have been impossible before practice. Let’s be clear, though, that this gain
comes at a price. With sufficient practice, task performance can go forward
with no executive control, and so the performance is essentially not controlled. This can create a setting in which the performance acts as a “mental
reflex,” going forward, once triggered, whether you like it or not.
Psychologists use the term automaticity to describe tasks that are well
practiced and involve little (or no) control. (For a classic statement, see
Shiffrin & Schneider, 1977; also Moors, 2016; Moors & De Houwer, 2006.)
The often-mentioned example is an effect known as Stroop interference. In
the classic demonstration of this effect, study participants are shown a series
of words and asked to name aloud the color of the ink used for each word.
The trick, though, is that the words themselves are color names. So people
Practice
•
185
might see the word “BLUE” printed in green ink and would have to say
“green” out loud, and so on (see Figure 5.21; Stroop, 1935).
This task turns out to be extremely difficult. There’s a strong tendency
to read the printed words themselves rather than to name the ink color, and
people make many mistakes in this task. Presumably, this reflects the fact
that word recognition, especially for college-age adults, is enormously well
practiced and therefore can proceed automatically. (For more on these issues,
including debate about what exactly automaticity involves, see Besner et al.,
2016; Durgin, 2000; Engle & Kane, 2004; Jacoby et al., 2003; Kane & Engle,
2003; Labuschagne & Besner, 2015; Moors, 2016.)
Where Are the Limits?
As we near the end of our discussion of attention, it may be useful again
to summarize where we are. Two simple ideas are key: First, tasks require
resources, and second, you can’t “spend” more resources than you have.
FIGURE 5.21
STROOP INTERFERENCE
Column A
Column B
As rapidly as you can, name out loud the colors of the ink in Column A.
(You’ll say, “black, green” and so on.) Next, do the same for Column B — again,
naming out loud the colors of the ink. You’ll probably find it much easier to
do this for Column A, because in Column B you experience interference from
the automatic habit of reading the words.
186 •
C H A P T E R F I V E Paying Attention
These claims are central for almost everything we’ve said about selective and
divided attention.
As we’ve seen, though, there are different types of resources, and the exact
resource demand of a task depends on several factors. The nature of the
task matters, of course, so that the resources required by a verbal task (e.g.,
reading) are different from those required by a spatial task (e.g., remembering a shape). The novelty of the task and the amount of flexibility the task
requires also matter. Connected to this, practice matters, with well-practiced
tasks requiring fewer resources.
What, then, sets the limits on divided attention? When can you do two
tasks at the same time, and when not? The answer varies, case by case. If two
tasks make competing demands on task-specific resources, the result will be
interference. If two tasks make competing demands on task-general resources
(the energy supply or executive control), again the result will be interference. Also, it will be especially difficult to combine tasks that involve similar stimuli — tasks that both involve printed text, for example, or that both
involve speech. The problem here is that these stimuli can “blur together,”
with a danger that you’ll lose track of which elements belong in which input
(“Was it the man who said ‘yes,’ or was it the woman?”; “Was the red dog in
the top picture or the bottom one?”). This sort of “crosstalk” (leakage of bits
of one input into the other input) can compromise your performance.
In short, it seems again like we need a multipart theory of attention, with
performance being limited by different factors at different times. This perspective draws us back to a claim we’ve made several times in this chapter:
Attention cannot be thought of as a skill or a mechanism. Instead, attention
is an achievement — an achievement of performing multiple activities simultaneously or an achievement of successfully avoiding distraction when you
want to focus on a single task. And this achievement rests on an intricate
base, so that many elements contribute to your ability to attend.
Finally, we have discussed various limits on human performance — that is,
limits on how much you can do at any one time. How rigid are these limits?
We’ve discussed the improvements in divided attention that are made possible by practice, but are there boundaries on what practice can accomplish?
Can you gain new mental resources or find new ways to accomplish a task
in order to avoid the bottleneck created by some limited resource? Some
evidence indicates that the answer may be yes; if so, many of the claims made
in this chapter must be understood as claims about what is usual, not about
what is possible. (See Hirst, Spelke, Reaves, Caharack, & Neisser, 1980;
Spelke, Hirst, & Neisser, 1976. For a neuroscience perspective, see Just &
Buchweitz, 2017.) With this, some traditions in the world — Buddhist meditation traditions, for example — claim it’s possible to train attention so that
one has better control over one’s mental life. How do these claims fit into
the framework we’ve developed in this chapter? These are issues in need of
exploration, and, in truth, what’s at stake here is a question about the boundaries on human potential, making these issues of deep interest for future
researchers to pursue.
TEST YOURSELF
11.Why does practice
decrease the resource
demand of a task?
12.Why does practice
create a situation in
which you can lose
control over your own
mental steps?
Practice
•
187
COGNITIVE PSYCHOLOGY AND EDUCATION
ADHD
When students learn about attention, they often have questions about failures
of attention: “Why can’t I focus when I need to?”; “Why am I so distracted
by my roommate moving around the room when I’m studying?”; “Why can
some people listen to music while they’re reading, but I can’t?”
One question comes up more than any other: “I [or “my friend” or “my
brother] was diagnosed with ADHD. What’s that all about?” This question
refers to a common diagnosis: attention-deficit/hyperactivity disorder. The disorder is characterized by a number of problems, including impulsivity, constant
fidgeting, and difficulty in keeping attention focused on a task. People with
ADHD hop from activity to activity and have trouble organizing or completing
projects. They sometimes have trouble following a conversation and are easily
distracted by an unimportant sight or sound. Even their own thoughts can
distract them — and so they can be pulled off track by their own daydreams.
The causes of ADHD are still unclear. Contributing factors that have been
mentioned include encephalitis, genetic influences, food allergies, high lead
concentrations in the environment, and more. The uncertainty about this
point comes from many sources, including some ambiguity in the diagnosis
of ADHD. There’s no question that there’s a genuine disorder, but diagnosis is
complicated by the fact that the disorder can vary widely in its severity. Some
people have relatively mild symptoms; others are massively disrupted, and
this variation can make diagnosis difficult.
In addition, some critics argue that in many cases the ADHD diagnosis
is just a handy label for children who are particularly active or who don’t
easily adjust to a school routine or a crowded classroom. Indeed, some critics
suggest that ADHD is often just a convenient categorization for physicians
or school counselors who don’t know how else to think about an especially
energetic child.
In cases in which the diagnosis is warranted, though, what does it involve? As we describe in the chapter, there are many steps involved in “paying
attention,” and some of those steps involve inhibition — so that we don’t follow
every stray thought, or every cue in the environment, wherever it may lead. For
most of us, this is no problem, and we easily inhibit our responses to most distractors. We’re thrown off track only by especially intrusive distractors — such
as a loud noise or a stimulus that has special meaning for us.
Some researchers propose, though, that people with ADHD have less
effective inhibitory circuits in their brains, making them more vulnerable to
momentary impulses and chance distractions. This is what leads to their scattered thoughts, their difficulty in schoolwork, and so on.
What can be done to help people with ADHD? One of the common treatments is Ritalin, a drug that is a powerful stimulant. It seems ironic that we’d
give a stimulant to people who are already described as too active and too
188 •
C H A P T E R F I V E Paying Attention
energetic, but the evidence suggests that Ritalin is effective in treating actual
cases of ADHD — plausibly because the drug activates the inhibitory circuits
within the brain, helping the person to guard against wayward impulses.
However, we probably shouldn’t rely on Ritalin as the sole treatment for
ADHD. One reason is the risk of overdiagnosis — it’s worrisome that this
drug may be routinely given to people, including young children, who don’t
actually have ADHD. Also, there are concerns about the long-term effects
and possible side effects of Ritalin, and this certainly motivates us to seek
other forms of treatment. (Common side effects include weight loss, insomnia, anxiety, and slower growth during childhood.) Some of the promising
alternatives involve restructuring of the environment. If people with ADHD
are vulnerable to distraction, we can help them by the simple step of reducing the sources of distraction in their surroundings. Likewise, if people with
ADHD are influenced by whatever cues they detect, we can surround them
with helpful cues — reminders of what they’re supposed to be doing and the
tasks they’re supposed to be working on. These simple interventions do seem
to be helpful, especially with adults diagnosed with ADHD.
Overall, then, our description of ADHD requires multiple parts. Researchers have considered a diverse set of causes, and there may be a diverse set
of psychological mechanisms involved in the disorder. (See pp. 168.) The
diagnosis probably is overused, but the diagnosis is surely genuine in many
cases, and the problems involved in ADHD are real and serious. Medication
can help, but we’ve noted the concern about side effects of the medication.
Environmental interventions can also help and may, in fact, be the best bet
for the long term, especially given the important fact that in most cases the
symptoms of ADHD diminish as the years go by.
For more on this topic . . .
Barkley, R. A. (2004). Adolescents with ADHD: An overview of empirically
based treatments. Journal of Psychiatric Practice, 10, 39–56.
Barkley, R. A. (2008). ADHD in adults: What the science says. New York, NY:
Guilford.
Zillmer, E. A., Spiers, M. V., & Culbertson, W. C. (2008). Principles of neuropsychology. Belmont, CA: Wadsworth.
Cognitive Psychology and Education
•
189
chapter review
SUMMARY
• People are often oblivious to unattended inputs;
they usually cannot tell if an unattended auditory
input was coherent prose or random words, and
they often do not detect unattended visual inputs,
even though such inputs are right in front of their
eyes. However, some aspects of the unattended
inputs are detected. For example, people can report
on the pitch of the unattended sound and whether
it contained human speech or some other sort
of noise. Sometimes they can also detect stimuli
that are especially meaningful; some people, for
example, hear their own name if it is spoken on the
unattended channel.
• These results suggest that perception may require the commitment of mental resources, with
some of these resources helping to prime the detectors needed for perception. This proposal is supported by studies of inattentional blindness — that
is, studies showing that perception is markedly
impaired if the perceiver commits no resources to
the incoming stimulus information. The proposal
is also supported by results showing that participants perceive more efficiently when they can anticipate the upcoming stimulus (and so can prime
the relevant detectors). In many cases, the anticipation is spatial — if, for example, participants
know that a stimulus is about to arrive at a particular location. This priming, however, seems to
draw on a limited-capacity system, with the result
that priming one stimulus or one position takes
away resources that might be spent on priming
some other stimulus.
• The ability to pay attention to certain regions
of space has caused many researchers to compare
190
attention to a spotlight beam, with the idea that
stimuli falling “within the beam” are processed
more efficiently than stimuli that fall “outside the
beam.” However, this spotlight analogy is potentially misleading. In many circumstances, people do
seem to devote attention to identifiable regions of
space, no matter what falls within those regions. In
other circumstances, attention seems to be objectbased, not space-based, and so people pay attention
to specific objects, not specific positions.
• Perceiving seems to require the commitment of
resources, and so do most other mental activities.
This observation suggests an explanation for the
limits on divided attention: It is possible to perform
two tasks simultaneously only if the two tasks do
not in combination demand more resources than
are available. Some of the relevant mental resources,
including executive control, are task-general, being
required in a wide variety of mental activities. Other
mental resources are task-specific, being required
only for tasks of a certain type.
• Divided attention is influenced by practice, with
the result that it is often easier to divide attention
between familiar tasks than between unfamiliar
tasks. In the extreme, practice may produce automaticity, in which a task seems to require virtually
no mental resources but is also difficult to control.
One proposal is that automaticity results from the
fact that decisions are no longer needed for a wellpracticed routine; instead, one can simply run off
the entire routine, doing on this occasion just what
one did on prior occasions.
KEY TERMS
selective attention (p. 150)
dichotic listening (p. 150)
attended channel (p. 150)
unattended channel (p. 150)
shadowing (p. 151)
filter (p. 152)
fixation target (p. 153)
inattentional blindness (p. 154)
change blindness (p. 155)
early selection hypothesis (p. 157)
late selection hypothesis (p. 157)
repetition priming (p. 160)
limited-capacity system (p. 164)
mental resources (p. 164)
spatial attention (p. 165)
endogenous control of attention (p. 172)
exogenous control of attention (p. 172)
unilateral neglect syndrome (p. 174)
divided attention (p. 177)
executive control (p. 181)
perseveration error (p. 182)
goal neglect (p. 182)
automaticity (p. 185)
Stroop interference (p. 185)
TEST YOURSELF AGAIN
1.What information do people reliably pick up
from the attended channel? What do they pick
up from the unattended channel?
2.How is inattentional blindness demonstrated?
What situations outside of the laboratory seem
to reflect inattentional blindness?
3.What evidence seems to confirm early selection?
What evidence seems to confirm late selection?
4.What are the differences between the way that
stimulus-based priming functions and the way
that expectation-based priming functions?
8.When you first start paying attention to an
input, your attention seems to be space-based.
Once you’ve learned a bit about the input,
though, your attention seems to be objectbased. How does this pattern fit with the idea
that you pay attention by anticipating the
input?
9.Why is it easier to divide attention between
very different activities (e.g., knitting while
listening to a lecture) than it is to divide
attention between more similar activities?
5.Why is there a “cost” associated with being
misled by expectation-based priming?
10.What is executive control, and why does it
create limits on your ability to divide your
attention between two simultaneous tasks?
6.In what ways does the notion of a spotlight
beam accurately reflect how spatial attention
functions?
11.Why does practice decrease the resource
demand of a task?
7.In what ways does the notion of a spotlight
beam differ from the way spatial attention
functions?
12.Why does practice create a situation in which
you can lose control over your own mental
steps?
191
THINK ABOUT IT
1. It’s easy to keep your attention focused on
materials that you understand. But if you try to
focus on difficult material, your mind is likely
to wander. Does the chapter help you in understanding why that is? Explain your response.
help those who do it become better at paying
attention — staying focused and not suffering
from distraction. Does the chapter help you
in understanding why that might be? Explain
your response.
2. People claim that some forms of meditation
training (including Buddhist meditation) can
E eBOOK DEMONSTRATIONS & ESSAYS
Go to http://digital.wwnorton.com/cognition7 for the online demonstrations and essays
relevant to this chapter. These can be found in the ebook.
Online Demonstrations
• Demonstration 5.1: Shadowing
• Demonstration 5.2: Color-Changing Card Trick
• Demonstration 5.3: The Control of Eye
Online Applying Cognitive Psychology and the
Law Essays
• Cognitive Psychology and the Law: Guiding the
Formulation of New Laws
Movements
• Demonstration 5.4: Automaticity and the
Stroop Effect
COGNITION LABS
Go to http://digital.wwnorton.com/cognition7 to experience interactive online psychology
labs relevant to this chapter.
192
Memory
3
part
A
s you move through life, you encounter new facts, gain new skills, and
have new experiences. And you’re often changed by all of this, so that
later on you know things and can do things that you couldn’t know or do
before. How do these changes happen? How do you get new information into your
memory, and then how do you retrieve this information when you need it? And
how much trust can you put in this process? Why, for example, do people some-
times not remember things (including important things)? And why are memories
sometimes wrong — so that in some cases you remember an event one way, but a
friend who was present at the same event remembers things differently?
We’ll tackle all these issues in this section, and they will lead us to theoretical claims and practical applications. We’ll offer suggestions, for example,
about how students should study their class materials to maximize retention,
and also what students can do later on so that they’ll retain things they learned
at some earlier point.
In our discussion, several themes will emerge again and again. One theme
concerns the active nature of learning, and we’ll discuss the fact that passive
exposure to information, with no intellectual engagement, leads to poor
memory. From this base, we’ll consider why some forms of engagement with
to-be-learned material lead to especially good memory but other forms do not.
A second theme concerns the role of memory connections. In Chapter 6,
we’ll see that learning involves the creation of connections, and the more connections formed, the better the learning. In Chapter 7, we’ll argue that these
connections can serve as “retrieval paths” — paths that, you hope, will lead you
from your memory search’s starting point to the information you’re trying to
recall. As we’ll see, this idea has clear implications for when you will remember
a previous event and when you won’t.
Chapter 8 then explores a different aspect of the connections idea: Memory
connections can actually be a source of memory errors. We’ll ask what this
means for memory accuracy overall, and we’ll discuss what you can do to minimize error and to improve the completeness and accuracy of your memory.
193
6
chapter
The Acquisition of
Memories and the
Working-Memory
System
what if…
Clive Wearing is an accomplished musician and a
scholar of Renaissance music. When he was 47 years
old, however, his brain was horribly damaged by a Herpes virus, and
he now has profound amnesia. He is still articulate and intelligent, able
to participate in an ongoing conversation, and still able to play music
(beautifully) and conduct. But ever since the viral infection, he’s been
unable to form new memories. He can’t recall any of the experiences
he’s had in the thirty years since he suffered the brain damage. He can’t
even remember events that happened just moments ago, with a bizarre
result: Every few minutes, Wearing realizes he can’t recall anything from
a few seconds back, and so he concludes that he must have just woken
up. He grabs his diary and writes “8:31 a.m. Now I am really, completely
awake.” A short while later, though, he again realizes he can’t recall the
last seconds, so he decides that now he has just woken up. He picks up
his diary to record this event and immediately sees his previous entry.
Puzzled, he crosses it out and replaces it with “9:06 a.m. Now I am perfectly, overwhelmingly awake.” But then the process repeats, and so this
entry, too, gets scribbled out and a new entry reads “9:34 a.m.: Now I am
superlatively, actually awake.”
Despite this massive disruption, Wearing’s love for his wife, Deborah,
has not in any way been diminished by his amnesia. But here, too,
the memory loss has powerful effects. Each time Deborah enters his
room — even if she’s been away just a few minutes — he races to embrace
her as though it’s been countless lonely months since they last met. If
asked directly, he has no recollection of her previous visits — including a
visit that might have happened just minutes earlier.
We met a different case of amnesia in Chapter 1 — the famous patient
H.M. He, too, was unable to recall his immediate past — and the many,
deep problems this produced included an odd sort of disorientation: If
you were smiling at him, was it because you’d just said something funny?
Or because he’d said something embarrassing? Or had you been smiling
all along? As H.M. put it, “Right now, I’m wondering. Have I done or said
anything amiss? You see, at this moment, everything looks clear to me,
but what happened just before? That’s what worries me” (Milner, 1970,
p. 37; also see Corkin, 2013).
Cases like these remind us how profoundly our memories shape our
everyday lives. But these cases also raise many questions: Why is it that
195
A
B
CLIVE WEARING
Clive Wearing (shown here with his wife) developed profound amnesia as a result of
viral encephalitis, and now he seems to have only a moment-to-moment consciousness. With no memory at all of what he was doing just seconds ago, he is often convinced he just woke up, and he repeatedly writes in his diary, “Now perfectly awake
(1st time).” On the diary page shown here, he has recorded his thought, at 5:42 a.m.,
as his “1st act” because he has no memory of any prior activity. Soon after, though,
he seems to realize again that he has no memory of any earlier events, and so he
scribbles out the entry and now records that his bath is his “1st act.” The sequence
repeats over and over, with Wearing never recalling what he did before his current
activity, and so he records act after act as his “first.”
196 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
preview of chapter themes
•
e begin the chapter with a discussion of the broad archiW
tecture of memory. We then turn to a closer examination
of one component of this architecture: working memory.
•
e emphasize the active nature of working memory —
W
activity that is especially evident when we discuss working memory’s “central executive,” a mental resource that
serves to order, organize, and control our mental lives.
•
he active nature of memory is also evident in the proT
cess of rehearsal. Rehearsal is effective only if the person
engages the materials in some way; this is reflected,
for example, in the contrast between deep processing
(which leads to excellent memory) and mere maintenance
rehearsal (which produces basically no memory benefit).
•
ctivity during learning appears to establish memory
A
connections, which can serve as retrieval routes when it
comes time to remember the target material. For complex
material, the best way to establish these connections is
to seek to understand the material; the better the understanding, the better the memory will be.
Wearing still remembers who his wife is? How is it possible that even
with his amnesia, Wearing remains such a talented musician? Why does
H.M. still remember his young adult years, even though he can’t remember what he said just five minutes earlier? We’ll tackle questions like
these in this chapter and the next two.
Acquisition, Storage, and Retrieval
How does new information — whether it’s a friend’s phone number or a fact
you hope to memorize for the bio exam — become established in memory?
Are there ways to learn that are particularly effective? Then, once information is in storage, how do you locate it and “reactivate” it later? And why
does search through memory sometimes fail — so that, for example, you
forget the name of that great restaurant downtown (but then remember the
name when you’re midway through a mediocre dinner someplace else)?
In tackling these questions, there’s a logical way to organize our inquiry.
Before there can be a memory, you need to gain, or “acquire,” some new
information. Therefore, acquisition — the process of gaining information
and placing it into memory — should be our first topic. Then, once you’ve
acquired this information, you need to hold it in memory until the information is needed. We refer to this as the storage phase. Finally, you remember. In
other words, you somehow locate the information in the vast warehouse that
is memory and you bring it into active use; this is called retrieval.
This organization seems logical; it fits, for example, with the way most
“electronic memories” (e.g., computers) work. Information (“input”) is provided to a computer (the acquisition phase). The information then resides
in some dormant form, generally on the hard drive or perhaps in the cloud
(the storage phase). Finally, the information can be brought back from this
dormant form, often via a search process that hunts through the disk (the
retrieval phase). And there’s nothing special about the computer comparison
here; “low-tech” information storage works the same way. Think about a file
Acquisition, Storage, and Retrieval
•
197
TEST YOURSELF
1.Define the terms
“acquisition,” “storage,”
and “retrieval.”
drawer — information is acquired (i.e., filed), rests in this or that folder, and
then is retrieved.
Guided by this framework, we’ll begin our inquiry by focusing on the
acquisition of new memories, leaving discussion of storage and retrieval for
later. As it turns out, though, we’ll soon find reasons for challenging this overall approach to memory. In discussing acquisition, for example, we might
wish to ask: What is good learning? What guarantees that material is firmly
recorded in memory? As we’ll see, evidence indicates that what counts as
“good learning” depends on how the memory is to be used later on, so that
good preparation for one kind of use may be poor preparation for a different kind of use. Claims about acquisition, therefore, must be interwoven
with claims about retrieval. These interconnections between acquisition and
retrieval will be the central theme of Chapter 7.
In the same way, we can’t separate claims about memory acquisition from
claims about memory storage. This is because how you learn (acquisition)
depends on what you already know (information in storage). We’ll explore
this important relationship in both this chapter and Chapter 8.
We begin, though, in this chapter, by describing the acquisition process.
Our approach will be roughly historical. We’ll start with a simple model,
emphasizing data collected largely in the 1970s. We’ll then use this as the
framework for examining more recent research, adding refinements to the
model as we proceed.
The Route into Memory
For many years, theorizing in cognitive psychology focused on the process
through which information was perceived and then moved into memory
storage — that is, on the process of information acquisition. One early proposal was offered by Waugh and Norman (1965). Later refinements were
added by Atkinson and Shiffrin (1968), and their version of the proposal
came to be known as the modal model. Figure 6.1 provides a simplified
depiction of this model.
Updating the Modal Model
According to the modal model, when information first arrives, it is stored
briefly in sensory memory. This form of memory holds on to the input in
“raw” sensory form — an iconic memory for visual inputs and an echoic
memory for auditory inputs. A process of selection and interpretation then
moves the information into short-term memory — the place where you hold
information while you’re working on it. Some of the information is then
transferred into long-term memory, a much larger and more permanent storage place.
This proposal captures some important truths, but it needs to be updated
in several ways. First, the idea of “sensory memory” plays a much smaller
role in modern theorizing, so modern discussions of perception (like our
198 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
FIGURE 6.1
N INFORMATION-PROCESSING VIEW
A
OF MEMORY
Retrieval
Sensory
memory
Incoming
information
Short-term
memory
Maintenance
via rehearsal
Long-term
memory
Lost
Diagrams like this one depict the flow of information hypothesized by the
modal model. The model captures many important truths but must be
updated in important ways. Current theorizing, for example, emphasizes that
short-term memory (now called “working memory”) is not a place serving as
a “loading dock” outside of long-term memory. Instead, working memory is
best understood as an activity, in ways described in the chapter.
ICONIC MEMORY
discussion in Chapters 2 and 3) often make no mention of this memory.
(For a recent assessment of visual sensory memory, though, see Cappiello
& Zhang, 2016.) Second, modern proposals use the term working memory
rather than “short-term memory,” to emphasize the function of this memory.
Ideas or thoughts in this memory are currently activated, currently being
thought about, and so they’re the ideas you’re currently working on. Longterm memory (LTM), in contrast, is the vast repository that contains all of
your knowledge and all of your beliefs — most of which you aren’t thinking
about (i.e., aren’t working on) at this moment.
The modal model also needs updating in another way. Pictures like the
one in Figure 6.1 suggest that working memory is a storage place, sometimes described as the “loading dock” just outside of the long-term memory
“warehouse.” The idea is that information has to “pass through” working
memory on the way into longer-term storage. Likewise, the picture implies
that memory retrieval involves the “movement” of information out of storage
and back into working memory.
In contrast, contemporary theorists don’t think of working memory as a
“place” at all. Instead, working memory is (as we will see) simply the name we
give to a status. Therefore, when we say that ideas are “in working memory,”
we simply mean that these ideas are currently activated and being worked on
by a specific set of operations.
In a classic experiment (Sperling,
1960), participants viewed a
grid like this one for just 50 ms.
If asked to report all of the
letters, participants could report just three or four of them.
In a second condition, participants saw the grid and then
immediately afterward heard
a cue signaling which row
they had to report. No matter which row they were asked
about, participants could recall most of the row’s letters.
It seems, therefore, that participants could remember the
entire display (in iconic memory) for a brief time, and could
“read off” the contents of any
row when appropriately cued.
The limitation in the report-all
condition, then, came from the
fact that iconic memory faded
away before the participants
could report on all of it.
The Route into Memory
•
199
We’ll have more to say about this modern perspective before we’re
through. It’s important to emphasize, though, that contemporary thinking
also preserves some key ideas from the modal model, including its claims
about how working memory and long-term memory differ from each other.
Let’s identify those differences.
First, working memory is limited in size; long-term memory is enormous.
In fact, long-term memory has to be enormous, because it contains all of
your knowledge — including specific knowledge (e.g., how many siblings you
have) and more general themes (e.g., that water is wet, that Dublin is in
Ireland, that unicorns don’t exist). Long-term memory also contains all of
your “episodic” knowledge — that is, your knowledge about events, including events early in your life as well as more recent experiences.
Second, getting information into working memory is easy. If you think
about a particular idea or some other type of content, then you’re “working
on” that idea or content, and so this information — by definition — is now in
your working memory. In contrast, we’ll see later in the chapter that getting
information into long-term memory often involves some work.
Third, getting information out of working memory is also easy. Since (by
definition) this memory holds the ideas you’re thinking about right now, the
information is already available to you. Finding information in long-term
memory, in contrast, can sometimes be difficult and slow — and in some
settings can fail completely.
Fourth, the contents of working memory are quite fragile. Working
memory, we emphasize, contains the ideas you’re thinking about right now.
If your thoughts shift to a new topic, therefore, the new ideas will enter
working memory, pushing out what was there a moment ago. Long-term
memory, in contrast, isn’t linked to your current thoughts, so it’s much less
fragile — information remains in storage whether you’re thinking about it
right now or not.
We can make all these claims more concrete by looking at some classic
research findings. These findings come from a task that’s quite artificial (i.e.,
not the sort of memorizing you do every day) but also quite informative.
Working Memory and Long-Term Memory:
One Memory or Two?
In many studies, researchers have asked participants to listen to a series of
words, such as “bicycle, artichoke, radio, chair, palace.” In a typical experiment, the list might contain 30 words and be presented at a rate of one word
per second. Immediately after the last word is read, the participants must
repeat back as many words as they can. They are free to report the words in
any order they choose, which is why this task is called a free recall procedure.
People usually remember 12 to 15 words in this test, in a consistent pattern.
They’re very likely to remember the first few words on the list, something
known as the primacy effect, and they’re also likely to remember the last
few words on the list, a recency effect. The resulting pattern is a U-shaped
200 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
100
Recency
effect
Percent recall
80
FIGURE 6.2 PRIMACY AND RECENCY
EFFECTS IN FREE RECALL
60
Primacy
effect
40
20
0
1
5
10
15
Serial position
20
Research participants in this study heard a list of 20 common words presented at a rate of one word per second.
Immediately after hearing the list, participants were asked to
write down as many of the words on the list as they could
recall. The results show that position in the series strongly
affected recall—participants had better recall for words at the
beginning of the list (the primacy effect) and for words at
the end of the list (the recency effect), compared to words
in the middle of the list.
curve describing the relation between positions within the series — or serial
position — and the likelihood of recall (see Figure 6.2; Baddeley & Hitch,
1977; Deese & Kaufman, 1957; Glanzer & Cunitz, 1966; Murdock, 1962;
Postman & Phillips, 1965).
Explaining the Recency Effect
What produces this pattern? We’ve already said that working memory contains the material someone is working on at just that moment. In other words,
this memory contains whatever the person is currently thinking about; and
during the list presentation, the participants are thinking about the words
they’re hearing. Therefore, it’s these words that are in working memory. This
memory, however, is limited in size, capable of holding only five or six words.
Consequently, as participants try to keep up with the list presentation, they’ll
be placing the words just heard into working memory, and this action will
bump the previous words out of working memory. As a result, as participants
proceed through the list, their working memories will, at each moment, contain only the half dozen words that arrived most recently. Any words that
arrived earlier than these will have been pushed out by later arrivals.
Of course, the last few words on the list don’t get bumped out of working memory, because no further input arrives to displace them. Therefore, when the list presentation ends, those last few words stay in place.
Moreover, our hypothesis is that materials in working memory are readily
available — easily and quickly retrieved. When the time comes for recall,
then, working memory’s contents (the list’s last few words) are accurately
and completely recalled.
The key idea, then, is that the list’s last few words are still in working
memory when the list ends (because nothing has arrived to push out these
items), and we know that working memory’s contents are easy to retrieve.
This is the source of the recency effect.
The Route into Memory
•
201
Explaining the Primacy Effect
The primacy effect has a different source. We’ve suggested that it takes
some work to get information into long-term memory (LTM), and it seems
likely that this work requires some time and attention. So let’s examine how
participants allocate their attention to the list items. As participants hear
the list, they do their best to be good memorizers, and so when they hear
the first word, they repeat it over and over to themselves (“bicycle, bicycle,
bicycle”) — a process known as memory rehearsal. When the second word
arrives, they rehearse it, too (“bicycle, artichoke, bicycle, artichoke”). Likewise for the third (“bicycle, artichoke, radio, bicycle, artichoke, radio”), and
so on through the list. Note, though, that the first few items on the list are
privileged. For a brief moment, “bicycle” is the only word participants have
to worry about, so it has 100% of their attention; no other word receives
this privilege. When “artichoke” arrives a moment later, participants divide
their attention between the first two words, so “artichoke” gets only 50%
of their attention — less than “bicycle” got, but still a large share of the
participants’ efforts. When “radio” arrives, it has to compete with “bicycle”
and “artichoke” for the participants’ time, and so it receives only 33% of
their attention.
Words arriving later in the list receive even less attention. Once six or
seven words have been presented, the participants need to divide their
attention among all these words, which means that each one receives only
a small fraction of the participants’ focus. As a result, words later in the
list are rehearsed fewer times than words early in the list — a fact that
can be confirmed simply by asking participants to rehearse out loud
(Rundus, 1971).
This view of things leads immediately to our explanation of the primacy
effect — that is, the observed memory advantage for the early list items.
These early words didn’t have to share attention with other words (because
the other words hadn’t arrived yet), so more time and more rehearsal were
devoted to them than to any others. This means that the early words have a
greater chance of being transferred into LTM — and so a greater chance of
being recalled after a delay. That’s what shows up in these classic data as the
primacy effect.
Testing Claims about Primacy and Recency
This account of the serial-position curve leads to many predictions. First, we’re
claiming the recency portion of the curve is coming from working memory,
while other items on the list are being recalled from LTM. Therefore, manipulations of working memory should affect recall of the recency items but not
items earlier in the list. To see how this works, consider a modification of our
procedure. In the standard setup, we allow participants to recite what they
remember immediately after the list’s end. But instead, we can delay recall
by asking participants to perform some other task before they report the list
items — for example, we can ask them to count backward by threes, starting
from 201. They do this for just 30 seconds, and then they try to recall the list.
202 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
We’ve hypothesized that at the end of the list working memory still contains the last few items heard from the list. But the task of counting backward will itself require working memory (e.g., to keep track of where you
are in the counting sequence). Therefore, this chore will displace working
memory’s current contents; that is, it will bump the last few list items out of
working memory. As a result, these items won’t benefit from the swift and
easy retrieval that working memory allows, and, of course, that retrieval was
the presumed source of the recency effect. On this basis, the simple chore of
counting backward, even if only for a few seconds, will eliminate the recency
effect. In contrast, the counting backward should have no impact on recall
of the items earlier in the list: These items are (by hypothesis) being recalled
from long-term memory, not working memory, and there’s no reason to
think the counting task will interfere with LTM. (That’s because LTM, unlike
working memory, isn’t dependent on current activity.)
Figure 6.3 shows that these predictions are correct. An activity interpolated,
or inserted, between the list and recall essentially eliminates the recency effect,
but it has no influence elsewhere in the list (Baddeley & Hitch, 1977; Glanzer
& Cunitz, 1966; Postman & Phillips, 1965). In contrast, merely delaying the
recall for a few seconds after the list’s end, with no interpolated activity, has
no impact. In this case, participants can continue rehearsing the last few items
during the delay and so can maintain them in working memory. With no new
materials coming in, nothing pushes the recency items out of working memory,
and so, even with a delay, a normal recency effect is observed.
100
30-second
unfilled
delay
90
Percentage of words recalled
80
Immediate
recall
70
60
50
30-second
filled
delay
40
30
20
10
0
1
2,3
4,5
6,7
Input position
8,9
10,11
12
FIGURE 6.3 THE IMPACT
OF INTERPOLATED ACTIVITY
ON THE RECENCY EFFECT
With immediate recall (the red line in
the figure), or if recall is delayed by
30 seconds with no activity during
the delay (the purple line), a strong
recency effect is detected. In contrast, if participants spend 30 seconds on some other activity between
hearing the list and the subsequent
memory test (the blue line), the
recency effect is eliminated. This interpolated activity has no impact on
the pre-recency portion of the curve
(i.e., the portion of the curve other
than the last few positions).
The Route into Memory
•
203
FIGURE 6.4 RATE OF LIST
PRESENTATION AND THE
SERIAL-POSITION EFFECT
Presenting the to-be-remembered materials at a slower rate improves pre-recency
performance but has no effect on recency.
The slow presentation rate in this case was
9 seconds per item; the faster rate was
3 seconds per item.
TEST YOURSELF
2.List the four ways in
which (either in the
modal model or in
more recent views)
working memory is
different from longterm storage.
3.How is the primacy
effect usually
explained? How is
the recency effect
usually explained?
204 •
Percentage of words recalled
100
75
Slow presentation
50
25
Fast presentation
1
5
10
15
20
Serial position
We’d expect a different outcome, though, if we manipulate long-term
memory rather than working memory. In this case, the manipulation should
affect all performance except for recency (which, again, is dependent on
working memory, not LTM). For example, what happens if we slow down the
presentation of the list? Now, participants will have more time to spend on
all of the list items, increasing the likelihood of transfer into more permanent
storage. This should improve recall for all items coming from LTM. Working memory, in contrast, is limited by its size, not by ease of entry or ease of
access. Therefore, the slower list presentation should have no influence on
working-memory performance. Research results confirm these claims: Slowing the list presentation improves retention of all the pre-recency items but
does not improve the recency effect (see Figure 6.4).
Other variables that influence long-term memory have similar effects.
Using more familiar or more common words, for example, would be expected to ease entry into long-term memory and does improve pre-recency
retention, but it has no effect on recency (Sumby, 1963).
It seems, therefore, that the recency and pre-recency portions of the
curve are influenced by distinct sets of factors and obey different principles.
Apparently, then, these two portions of the curve are the products of different mechanisms, just as our theory proposed. In addition, fMRI scans suggest
that memory for early items on a list depends on brain areas (in and around
the hippocampus) that are associated with long-term memory; memory
for later items on the list do not show this pattern (Talmi, Grady, GoshenGottstein, & Moscovitch, 2005; also Eichenbaum, 2017; see Figure 6.5). This
provides further confirmation for our memory model.
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
FIGURE 6.5
RAIN REGIONS SUPPORTING WORKING
B
MEMORY AND LONG-TERM MEMORY
Retrieval from long-term
memory specifically
activates the hippocampus.
Retrieval from working
memory specifically activates
the perirhinal cortex.
We can confirm the distinction between working memory and long-term
memory with fMRI scans. These scans suggest that memory for early items
on a list depends on brain areas (in and around the hippocampus) that are
associated with long-term memory; memory for later items on the list do not
show this pattern.
(talmi, grady, goshen-gottstein, & moscovitch, 2005)
A Closer Look at Working Memory
Earlier, we counted four fundamental differences between working memory and LTM — the size of these two stores, the ease of entry, the ease of
retrieval, and the fact that working memory is dependent on current activity
(and therefore fragile) while LTM is not. These are all points proposed by the
modal model and preserved in current thinking. As we’ve said, though, investigators’ understanding of working memory has developed over the years.
Let’s examine the newer conception in more detail.
The Function of Working Memory
Virtually all mental activities require the coordination of several pieces of
information. Sometimes the relevant bits come into view one by one, so that
you need to hold on to the early-arrivers until the rest of the information is
available, and only then weave all the bits together. Alternatively, sometimes
the relevant bits are all in view at the same time — but you still need to hold
on to them together, so that you can think about the relations and combinations. In either case, you’ll end up with multiple ideas in your thoughts, all
A Closer Look at Working Memory
•
205
activated simultaneously, and thus several bits of information in the status we
describe as “in working memory.” (For more on how you manage to focus on
these various bits, see Oberauer & Hein, 2012.)
Framing things in this way makes it clear how important working memory is: You use it whenever you have multiple ideas in your mind, multiple
elements that you’re trying to combine or compare. Let’s now add that
people differ in the “holding capacity” of their working memories. Some
people are able to hold on to (and work with) more elements, and some
with fewer. How does this matter? To find out, we first need a way of
measuring working memory’s capacity, to determine if your memory capacity is above average, below, or somewhere in between. The procedure for
obtaining this measurement, however, has changed over the years; looking
at this change will help clarify what working memory is, and what working
memory is for.
Digit Span
For many years, the holding capacity of working memory was measured with
a digit-span task. In this task, research participants hear a series of digits read
to them (e.g., “8, 3, 4”) and must immediately repeat them back. If they do so
successfully, they’re given a slightly longer list (e.g., “9, 2, 4, 0”). If they can
repeat this one without error, they’re given a still longer list (“3, 1, 2, 8, 5”),
and so on. The procedure continues until the participant starts to make
errors — something that usually happens when the list contains more than
seven or eight items. The number of digits the person can echo back without
errors is referred to as that person’s digit span.
Procedures such as this imply that working memory’s capacity is typically around seven items — at least five and probably not more than nine.
These estimates have traditionally been summarized by the statement that
this memory holds “7 plus-or-minus 2” items (Chi, 1976; Dempster, 1981;
Miller, 1956; Watkins, 1977).
However, we immediately need a refinement of these measurements.
If working memory can hold 7 plus-or-minus 2 items, what exactly is an
“item”? Can people remember seven sentences as easily as seven words?
Seven letters as easily as seven equations? In a classic paper, George Miller
(one of the founders of the field of cognitive psychology) proposed that
working memory holds 7 plus-or-minus 2 chunks (Miller, 1956). The term
“chunk” doesn’t sound scientific or technical, and that’s useful because this
informal terminology reminds us that a chunk doesn’t hold a fixed quantity
of information. Instead, Miller proposed, working memory holds 7 plusor-minus 2 packages, and what those packages contain is largely up to the
individual person.
The flexibility in how people “chunk” input can easily be seen in the span
test. Imagine that we test someone’s “letter span” rather than their “digit
span,” using the procedure already described. So the person might hear
206 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
“R, L” and have to repeat this sequence back, and then “F, C, H,” and so on.
Eventually, let’s imagine that the person hears a much longer list, perhaps
one starting “H, O, P, T, R, A, S, L, U. . . .” If the person thinks of these as
individual letters, she’ll only remember 7 of them, more or less. But she
might reorganize the list into “chunks” and, in particular, think of the letters as forming syllables (“HOP, TRA, SLU, . . .”). In this case, she’ll still
remember 7 plus-or-minus 2 items, but the items are syllables, and by remembering the syllables she’ll be able to report back at least a dozen letters
and probably more.
How far can this process be extended? Chase and Ericsson (1982;
Ericsson, 2003) studied a remarkable individual who happens to be a fan of
track events. When he hears numbers, he thinks of them as finishing times
for races. The sequence “3, 4, 9, 2,” for example, becomes “3 minutes and
49.2 seconds, near world-record mile time.” In this way, four digits become
one chunk of information. This person can then retain 7 finishing times
(7 chunks) in memory, and this can involve 20 or 30 digits! Better still, these
chunks can be grouped into larger chunks, and these into even larger chunks.
For example, finishing times for individual racers can be chunked together
into heats within a track meet, so that, now, 4 or 5 finishing times (more than
a dozen digits) become one chunk. With strategies like this and a lot of practice, this person has increased his apparent memory span from the “normal”
7 digits to 79 digits.
However, let’s be clear that what has changed through practice is merely
this person’s chunking strategy, not the capacity of working memory itself.
This is evident in the fact that when tested with sequences of letters, rather
than numbers, so that he can’t use his chunking strategy, this individual’s
memory span is a normal size — just 6 consonants. Thus, the 7-chunk limit
is still in place for this man, even though (with numbers) he’s able to make
extraordinary use of these 7 slots.
Operation Span
Chunking provides one complication in our measurement of working memory’s capacity. Another — and deeper — complication grows out of the very
nature of working memory. Early theorizing about working memory, as we
said, was guided by the modal model, and this model implies that working
memory is something like a box in which information is stored or a location
in which information can be displayed. The traditional digit-span test fits
well with this idea. If working memory is like a box, then it’s sensible to ask
how much “space” there is in the box: How many slots, or spaces, are there
in it? This is precisely what the digit span measures, on the idea that each
digit (or each chunk) is placed in its own slot.
We’ve suggested, though, that the modern conception of working memory is more dynamic — so that working memory is best thought of as a
status (something like “currently activated”) rather than a place. (See, e.g.,
A Closer Look at Working Memory
•
207
FIGURE 6.6
IS WORKING MEMORY A “PLACE”?
Verbal and numbers
Objects
Spatial
Problem solving
Modern theorists argue that working memory is not a place at all, but is
instead the name we give for a certain set of mental activities. Consistent with
this modern view, there’s no specific location within the brain that serves as
working memory. Instead, working memory is associated with a wide range
of brain sites, as shown here.
(after cabeza & nyberg, 2000)
Christophel, Klink, Spitzer, Roelfsema, & Haynes, 2017; also Figure 6.6.)
On this basis, perhaps we need to rethink how we measure this memory’s
capacity — seeking a measure that reflects working memory’s active
operation.
Modern researchers therefore measure this memory’s capacity in terms
of operation span, a measure of working memory when it is “working.”
There are several ways to measure operation span, with the types differing in what “operation” they use (e.g., Bleckley, Foster, & Engle, 2015;
Chow & Conway, 2015). One type is reading span. To measure this span,
a research participant might be asked to read aloud a series of sentences,
like these:
Due to his gross inadequacies, his position as director was terminated
abruptly.
It is possible, of course, that life did not arise on Earth at all.
Immediately after reading the sentences, the participant is asked to recall
each sentence’s final word — in this case, “abruptly” and “all.” If she can
do this with these two sentences, she’s asked to do the same task with a
group of three sentences, and then with four, and so on, until the limit on
208 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
FIGURE 6.7
DYNAMIC MEASURES OF WORKING MEMORY
(7
7) + 1 = 50; dog
(10/2) + 6 = 10; gas
(4
2) + 1 = 9; nose
(3/1) + 1 = 5; beat
(5/5) + 1 = 2; tree
Operation span can be measured in several different ways. In one procedure,
participants must announce whether each of these “equations” is true or
false, and then recall the words appended to each equation. If participants
can do this with two equations, we ask them to do three; if they can do that,
we ask them to try four. By finding out how far they can go, we measure their
working-memory capacity.
her performance is located. This limit defines the person’s working-memory
capacity, or WMC. (However there are other ways to measure operation
span — see Figure 6.7.)
Let’s think about what this task involves: storing materials (the ending
words) for later use in the recall test, while simultaneously working with
other materials (the full sentences). This juggling of processes, as the participant moves from one part of the task to the next, is exactly what working
memory must do in day-to-day life. Therefore, performance in this test is
likely to reflect the efficiency with which working memory will operate in
more natural settings.
Is operation span a valid measure — that is, does it measure what it’s supposed to? Our hypothesis is that someone with a higher operation span has
a larger working memory. If this is right, then someone with a higher span
should have an advantage in tasks that make heavy use of this memory.
Which tasks are these? They’re tasks that require you to keep multiple ideas
active at the same time, so that you can coordinate and integrate various
bits of information. So here’s our prediction: People with a larger span (i.e.,
a greater WMC) should do better in tasks that require the coordination of
different pieces of information.
Consistent with this claim, people with a greater WMC do have an advantage in many settings — in tests of reasoning, assessments of reading comprehension, standardized academic tests (including the verbal SAT), tasks that
require multitasking, and more. (See, e.g., Ackerman, Beier, & Boyle, 2002;
Butler, Arrington, & Weywadt, 2011; Daneman & Hannon, 2001; Engle &
Kane, 2004; Gathercole & Pickering, 2000; Gray, Chabris, & Braver, 2003;
Redick et al., 2016; Salthouse & Pink, 2008. For some complications, see
A Closer Look at Working Memory
•
209
Chow & Conway, 2015; Harrison, Shipstead, & Engle, 2015; Kanerva &
Kalakoski, 2016; Mella, Fagot, Lecert, & de Ribaupierre, 2015.)
These results convey several messages. First, the correlations between
WMC and performance provide indications about when it’s helpful to have
a larger working memory, which in turn helps us understand when and how
working memory is used. Second, the link between WMC and measures of
intellectual performance provide an intriguing hint about what we’re measuring with tests (like the SAT) that seek to measure “intelligence.” We’ll return
to this issue in Chapter 13 when we discuss the nature of intelligence. Third,
it’s important that the various correlations are observed with the more active
measure of working memory (operation span) but not with the more traditional (and more static) span measure. This point confirms the advantage of
the more dynamic measures and strengthens the idea that we’re now thinking
about working memory in the right way: not as a passive storage box, but
instead as a highly active information processor.
The Rehearsal Loop
Working memory’s active nature is also evident in another way: in the actual
structure of this memory. The key here is that working memory is not a single
entity but is instead, a system built of several components (Baddeley, 1986,
1992, 2012; Baddeley & Hitch, 1974; also see Logie & Cowan, 2015). At
the center of the working-memory system is a set of processes we discussed
in Chapter 5: the executive control processes that govern the selection and
sequence of thoughts. In discussions of working memory, these processes
have been playfully called the “central executive,” as if there were a tiny
agent embedded in your mind, running your mental operations. Of course,
there is no agent, and the central executive is just a name we give to the set of
mechanisms that do run the show.
The central executive is needed for the “work” in working memory; if you
have to plan a response or make a decision, these steps require the executive.
But in many settings, you need less than this from working memory. Specifically, there are settings in which you need to keep ideas in mind, not because
you’re analyzing them right now but because you’re likely to need them soon.
In this case, you don’t need the executive. Instead, you can rely on the executive’s “helpers,” leaving the executive free to work on more difficult matters.
Let’s focus on one of working memory’s most important helpers, the
articulatory rehearsal loop. To see how the loop functions, try reading the
next few sentences while holding on to these numbers: “1, 4, 6, 3.” Got them?
Now read on. You’re probably repeating the numbers over and over to yourself, rehearsing them with your inner voice. But this takes very little effort, so
you can continue reading while doing this rehearsal. Nonetheless, the moment
you need to recall the numbers (what were they?), they’re available to you.
In this setting, the four numbers were maintained by working memory’s
rehearsal loop, and with the numbers thus out of the way, the central executive could focus on the processes needed for reading. That is the advantage of
210 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
this system: With mere storage handled by the helpers, the executive is available for other, more demanding tasks.
To describe this sequence of events, researchers would say that you used
subvocalization — silent speech — to launch the rehearsal loop. This production by the “inner voice” produced a representation of the target numbers in
the phonological buffer, a passive storage system used for holding a representation (essentially an “internal echo”) of recently heard or self-produced
sounds. In other words, you created an auditory image in the “inner ear.” This
image started to fade away after a second or two, but you then subvocalized
the numbers once again to create a new image, sustaining the material in this
buffer. (For a glimpse of the biological basis for the “inner voice” and “inner
ear,” see Figure 6.8.)
FIGURE 6.8
B
RAIN ACTIVITY AND WORKING-MEMORY
REHEARSAL
Verbal
memory
Spatial
memory
Left lateral
Superior
Right lateral
Color is used here as an indication of increased brain activity (measured
in this case by positron emission tomography). When research participants
are doing a verbal memory task (and using the articulatory loop), activation
increases in areas ordinarily used for language production and perception.
A very different pattern is observed when participants are doing a task requiring memory for spatial position. Notice, then, that the “inner voice” and
“inner ear” aren’t casual metaphors; instead, they involve mechanisms that
are ordinarily used for overt speech and actual hearing.
( after jonides , lacey , & nee , 2005; also see jonides et al ., 2008)
A Closer Look at Working Memory
•
211
Many lines of evidence confirm this proposal. For example, when people
are storing information in working memory, they often make “sound-alike”
errors: Having heard “F,” they’ll report back “S.” When trying to remember the name “Tina,” they’ll slip and recall “Deena.” The problem isn’t that
people mis-hear the inputs at the start; similar sound-alike confusions emerge
if the inputs are presented visually. So, having seen “F,” people are likely to
report back “S”; they aren’t likely in this situation to report back the similarlooking “E.”
What produces this pattern? The cause lies in the fact that for this task
people are relying on the rehearsal loop, which involves a mechanism (the
“inner ear”) that stores the memory items as (internal representations of)
sounds. It’s no surprise, therefore, that errors, when they occur, are shaped by
this mode of storage.
As a test of this claim, we can ask people to take the span test while
simultaneously saying “Tah-Tah-Tah” over and over, out loud. This concurrent
articulation task obviously requires the mechanisms for speech production. Therefore, those mechanisms are not available for other use, including subvocalization. (If you’re directing your lips and tongue to produce the
“Tah-Tah-Tah” sequence, you can’t at the same time direct them to produce
the sequence needed for the subvocalized materials.)
How does this constraint matter? First, note that our original span test
measured the combined capacities of the central executive and the loop. That
is, when people take a standard span test (as opposed to the more modern
measure of operation span), they store some of the to-be-remembered items
in the loop and other items via the central executive. (This is a poor use of the
executive, underutilizing its talents, but that’s okay here because the standard
span task doesn’t require anything beyond mere storage.)
With concurrent articulation, though, the loop isn’t available for use, so
we’re now measuring the capacity of working memory without the rehearsal
loop. We should predict, therefore, that concurrent articulation, even though
it’s extremely easy, should cut memory span drastically. This prediction turns
out to be correct. Span is ordinarily about seven items; with concurrent
articulation, it drops by roughly a third — to four or five items (Chincotta &
Underwood, 1997; see Figure 6.9).
Second, with visually presented items, concurrent articulation should
eliminate the sound-alike errors. Repeatedly saying “Tah-Tah-Tah” blocks
use of the articulatory loop, and it’s in this loop, we’ve proposed, that the
sound-alike errors arise. This prediction, too, is correct: With concurrent
articulation and visual presentation of the items, sound-alike errors are
largely eliminated.
The Working-Memory System
As we have mentioned, your working memory contains the thoughts and
ideas you’re working on right now, and often this means you’re trying to
keep multiple ideas in working memory all at the same time. That can cause
212 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
Control
Suppression
10
Digit span (no. of items)
9
8
7
6
5
4
Chinese
English
Finnish
Greek
Language
Spanish
Swedish
FIGURE 6.9 THE EFFECT
OF CONCURRENT
ARTICULATION ON SPAN
In the Control condition, participants
were given a normal digit-span
test. In the Suppression condition, participants were required
to do concurrent articulation
while taking the test. Concurrent
articulation is easy, but it blocks
use of the articulatory loop and
consistently decreases memory
span, from roughly seven items
to five or so. And, plainly, this
use of the articulatory loop is
not an occasional strategy; instead, it can be found in a wide
range of countries and languages.
(after chincotta & underwood, 1997)
difficulties, because working memory only has a small capacity. That’s why
working memory’s helpers are so important, because they substantially
increase working memory’s capacity.
Against this backdrop, it’s not surprising that the working-memory system relies on other helpers in addition to the rehearsal loop. For example,
the system also relies on the visuospatial buffer, used for storing visual
materials such as mental images, in much the same way that the rehearsal
loop stores speech-based materials. (We’ll have more to say about mental
images in Chapter 11.) Baddeley (the researcher who launched the idea of a
working-memory system) has also proposed another component of the system: the episodic buffer. This component is proposed as a mechanism that
helps the executive organize information into a chronological sequence — so
that, for example, you can keep track of a story you’ve just heard or a film
clip you’ve just seen (e.g., Baddeley, 2000, 2012; Baddeley & Wilson, 2002;
Baddeley, Eysenck, & Anderson, 2009). The role of this component is evident
in patients with profound amnesia who seem unable to put new information into long-term storage, but who still can recall the flow of narrative in
a story they just heard. This short-term recall, it seems, relies on the episodic
buffer — an aspect of working memory that’s unaffected by the amnesia.
In addition, other helpers can be documented in some groups of people.
Consider people who have been deaf since birth and communicate via sign
language. We wouldn’t expect these individuals to rely on an “inner voice”
and an “inner ear” — and they don’t. People who have been deaf since birth
A Closer Look at Working Memory
•
213
rely on a different helper for working memory: They use an “inner hand”
(and covert sign language) rather than an “inner voice” (and covert speech).
As a result, they are disrupted if they’re asked to wiggle their fingers during
a memory task (similar to a hearing person saying “Tah-Tah-Tah”), and they
also tend to make “same hand-shape” errors in working memory (similar to
the sound-alike errors made by the hearing population).
The Central Executive
TEST YOURSELF
4.What does it mean
to say that working
memory holds seven
(plus-or-minus two)
“chunks”? What is a
chunk?
5.What evidence suggests that operation span is a better
measure of working
memory than the
more standard digitspan measure?
6.How does the rehearsal loop manage
to hold on to information with only occasional involvement by
the central executive?
What can we say about the main player within the working-memory
system — the central executive? In our discussion of attention (in Chapter 5),
we argued that executive control processes are needed to govern the sequence
of thoughts and actions; these processes enable you to set goals, make plans for
reaching those goals, and select the steps needed for implementing those plans.
Executive control also helps whenever you want to rise above habit or routine,
in order to “tune” your words or deeds to the current circumstances.
For purposes of the current chapter, though, let’s emphasize that the same
processes control the selection of ideas that are active at any moment in time.
And, of course, these active ideas (again, by definition) constitute the contents of working memory. It’s inevitable, then, that we would link executive
control with this type of memory.
With all these points in view, we’re ready to move on. We’ve now updated the modal model (Figure 6.1) in important ways, and in particular
we’ve abandoned the notion of a relatively passive short-term memory serving largely as storage container. We’ve shifted to a dynamic conception of
working memory, with the proposal that this term is merely the name for an
organized set of activities — especially the complex activities of the central
executive together with its various helpers.
But let’s also emphasize that in this modern conception, just as in the
modal model, working memory is quite fragile. Each shift in attention brings
new information into working memory, and the newly arriving material
displaces earlier items. Storage in this memory, therefore, is temporary.
Obviously, then, we also need some sort of enduring memory storage, so that
we can remember things that happened an hour, or a day, or even years ago.
Let’s turn, therefore, to the functioning of long-term memory.
Entering Long-Term Storage:
The Need for Engagement
We’ve already seen an important clue regarding how information gets established in long-term storage: In discussing the primacy effect, we suggested
that the more an item is rehearsed, the more likely you are to remember
that item later. To pursue this point, though, we need to ask what exactly
rehearsal is and how it might work to promote memory.
214 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
Two Types of Rehearsal
The term “rehearsal” doesn’t mean much beyond “thinking about.” In other
words, when a research participant rehearses an item on a memory list, she’s
simply thinking about that item — perhaps once, perhaps over and over; perhaps
mechanically, or perhaps with close attention to what the item means. Therefore, there’s considerable variety within the activities that count as rehearsal,
and psychologists find it useful to sort this variety into two broad types.
As one option, people can engage in maintenance rehearsal, in which they
simply focus on the to-be-remembered items themselves, with little thought
about what the items mean or how they relate to one another. This is a rote,
mechanical process, recycling items in working memory by repeating them
over and over. In contrast, relational, or elaborative, rehearsal involves thinking about what the to-be-remembered items mean and how they’re related to
one another and to other things you already know.
Relational rehearsal is vastly superior to maintenance rehearsal for
establishing information in memory. In fact, in many settings maintenance
rehearsal provides no long-term benefit at all. As an informal demonstration
of this point, consider the following experience (although, for a formal demonstration of this point, see Craik & Watkins, 1973). You’re watching your
favorite reality show on TV. The announcer says, “To vote for Contestant #4,
text 4 to 21523 from your mobile phone!” You reach into your pocket for
your phone but realize you left it in the other room. So you recite the number
to yourself while scurrying for your phone, but then, just before you dial,
you see that you’ve got a text message. You pause, read the message, and
then you’re ready to dial, but . . . you don’t have a clue what the number was.
What went wrong? You certainly heard the number, and you rehearsed it a
couple of times while moving to grab your phone. But despite these rehearsals,
the brief interruption from reading the text message seems to have erased the
number from your memory. However, this isn’t ultra-rapid forgetting. Instead,
you never established the number in memory in the first place, because in
this setting you relied only on maintenance rehearsal. That kept the number
in your thoughts while you were moving across the room, but it did nothing
to establish the number in long-term storage. And when you try to dial the
number after reading the text message, it’s long-term storage that you need.
The idea, then, is that if you think about something only in a mindless and mechanical way, the item won’t be established in your long-term
memory. Similarly, long-lasting memories aren’t created simply by repeated
exposures to the items to be remembered. If you encounter an item over and
over but, at each encounter, barely think about it (or think about it only in
a mechanical way), then this, too, won’t produce a long-term memory. As
a demonstration, consider the ordinary penny. Adults in the United States
have probably seen pennies tens of thousands of times. Adults in other
countries have seen their own coins just as often. If sheer exposure is what
counts for memory, people should remember perfectly what these coins
look like.
WE DON’T REMEMBER
THINGS WE DON’T
PAY ATTENTION TO
To promote public safety,
many buildings have fire
extinguishers and automatic
defibrillators positioned in
obvious and easily accessible
locations. But in a moment of
need, will people in the building remember where this
safety equipment is located?
Will they even remember
that the safety equipment
is conveniently available?
Research suggests they may
not. Occupants of the building have passed by the safety
equipment again and again—
but have had no reason to
notice the equipment. As a
result, they’re unlikely to remember where the equipment is located. (After Castel,
Vendetti, & Holyoak, 2012)
Entering Long-Term Storage: The Need for Engagement
•
215
But, of course, most people have little reason to pay attention to the penny.
Pennies are a different color from the other coins, so they can be identified
at a glance without further scrutiny. And, if it’s scrutiny that matters for
memory — or, more broadly, if we remember what we pay attention to and
think about — then memory for the coin should be quite poor.
The evidence on this point is clear: People’s memory for the penny is
remarkably bad. For example, most people know that Lincoln’s head is on
the “heads” side, but which way is he facing? Is it his right cheek that’s visible
or his left? What other markings are on the coin? Most people do very badly
with these questions; their answers to the “Which way is he facing?” question
are close to random (Nickerson & Adams, 1979). And performance is similar
for people in other countries remembering their own coins. (Also see Bekerian
& Baddeley, 1980; Rinck, 1999, for a much more consequential example.)
As a related example, consider the logo that identifies Apple products — the
iPhone, the iPad, or one of the Apple computers. Odds are good that you’ve
seen this logo hundreds and perhaps thousands of time, but you’ve probably had no reason to pay attention to its appearance. The prediction, then,
is that your memory for the logo will be quite poor — and this prediction is
correct. In one study, only 1 of 85 participants was able to draw the logo
correctly — with the bite on the proper side, the stem tilted the right way, and
the dimple properly placed in the logo’s bottom (Blake, Nazarian, & Castel,
2015; see Figure 6.10). And — surprisingly — people who use an Apple
computer (and therefore see the logo every time they turn on the machine)
perform at a level not much better than people who use a PC.
The Need for Active Encoding
Apparently, it takes some work to get information into long-term memory.
Merely having an item in front of your eyes isn’t enough — even if the item
is there over and over and over. Likewise, repeatedly thinking about an item
doesn’t, by itself, establish a memory. That’s evident in the fact that maintenance rehearsal seems ineffective at promoting memory.
FIGURE 6.10 MEMORY FOR AN
OFTEN-VIEWED LOGO
Most people have seen the Apple logo
countless times, but they’ve had no reason to
pay attention to its features. As a result, they
have poor memories for the features. Test
yourself. Can you find the correct version
among the options displayed here?
(the answer is at the end of the chapter.)
216 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
Further support for these claims comes from studies of brain activity during learning. In several procedures, researchers have used fMRI recording to
keep track of the moment-by-moment brain activity in participants who were
studying a list of words (Brewer, Zhao, Desmond, Glover, & Gabrieli, 1998;
Wagner, Koutstaal, & Schacter, 1999; Wagner et al., 1998; also see Levy,
Kuhl, & Wagner, 2010). Later, the participants were able to remember some
of the words they had learned, but not others, which allowed the investigators to return to their initial recordings and compare brain activity during the
learning process for words that were later remembered and words that were
later forgotten. Figure 6.11 shows the results, with a clear difference, during
the initial encoding, between these two types of words. Greater levels of brain
FIGURE 6.11
BRAIN ACTIVITY DURING LEARNING
Learn a series of words,
and, during learning,
record the neural
response to each word.
Based on what happened
at Time 2, go back and
examine the data from
Time 1, looking separately
at what happened during
learning for words that
were later remembered,
and what happened during
learning for words
that were later forgotten.
Test memory for
the words.
A
Left medial temporal lobe
Left inferior prefrontal cortex
Remembered
Forgotten
3
2
1
0
–1
0
B
Remembered
Forgotten
4
Activity level
Activity level
4
3
2
1
0
–1
4
Time (s)
8
12
0
4
8
Time (s)
12
(Panel A) Participants in this study were given a series of words to memorize, and their brain activity was
recorded during this initial presentation. These brain scans were then divided into two types: those showing
brain activity during the encoding of words that were remembered in the subsequent test, and those showing
activity during encoding of words that were forgotten in the test. (Panel B) As the figure shows, activity levels
during encoding were higher for the later-remembered words than they were for the later-forgotten words. This
finding confirms that whether a word is forgotten or not depends on participants’ mental activity when they
encountered the word in the first place.
Entering Long-Term Storage: The Need for Engagement
•
217
activity (especially in the hippocampus and regions of the prefrontal cortex)
were reliably associated with greater probabilities of retention later on.
These fMRI results are telling us, once again, that learning is not a passive process. Instead, activity is needed to lodge information into long-term
memory, and, apparently, higher levels of this activity lead to better memory. But this raises some new questions: What is this activity? What does it
accomplish? And if — as it seems — maintenance rehearsal is a poor way to
memorize, what type of rehearsal is more effective?
Incidental Learning, Intentional Learning,
and Depth of Processing
Consider a student taking a course in college. The student knows that her
memory for the course materials will be tested later (e.g., in the final exam).
And presumably she’ll take various steps to help herself remember: She may
read through her notes again and again; she may discuss the material with
friends; she may try outlining the material. Will these various techniques
work — so that she’ll have a complete and accurate memory when the exam
takes place? And notice that the student is taking these steps in the context
of wanting to memorize; she wants to do well on the exam! How does this
motivation influence performance? In other words, how does the intention to
memorize influence how or how well material is learned?
In an early experiment, participants in one condition heard a list of
24 words; their task was to remember as many as they could. This is
intentional learning — learning that is deliberate, with an expectation that
memory will be tested later. Other groups of participants heard the same
24 words but had no idea that their memories would be tested. This allows us
to examine the impact of incidental learning — that is, learning in the absence
of any intention to learn. One of the incidental-learning groups was asked
simply, for each word, whether the word contained the letter e. A different
incidental-learning group was asked to look at each word and to report how
many letters it contained. Another group was asked to consider each word
and to rate how pleasant it seemed.
Later, all the participants were tested — and asked to recall as many of the
words as they could. (The test was as expected for the intentional-learning
group, but it was a surprise for the other groups.) The results are shown in
Figure 6.12A (Hyde & Jenkins, 1969). Performance was relatively poor for
the “Find the e” and “Count the letters” groups but appreciably better for
the “How pleasant?” group. What’s striking, though, is that the “How pleasant?” group, with no intention to memorize, performed just as well as the
intentional-learning (“Learn these!”) group. The suggestion, then, is that the
intention to learn doesn’t add very much; memory can be just as good without
this intention, provided that you approach the materials in the right way.
This broad pattern has been reproduced in many other experiments (to
name just a few: Bobrow & Bower, 1969; Craik & Lockhart, 1972; Hyde &
Jenkins, 1973; Jacoby, 1978; Lockhart, Craik, & Jacoby, 1976; Parkin, 1984;
218 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
Slamecka & Graf, 1978). As one example, consider a study by Craik and
Tulving (1975). Their participants were led to do incidental learning (i.e.,
they didn’t know their memories would be tested). For some of the words
shown, the participants did shallow processing — that is, they engaged the
material in a superficial way. Specifically, they had to say whether the word
was printed in CAPITAL letters or not. (Other examples of shallow processing would be decisions about whether the words are printed in red or
in green, high or low on the screen, etc.) For other words, the participants
had to do a moderate level of processing: They had to judge whether each
word shown rhymed with a particular cue word. Finally, for other words,
participants had to do deep processing. This is processing that requires some
thought about what the words mean; specifically, Craik and Tulving asked
whether each word shown would fit into a particular sentence.
The results are shown in Figure 6.12B. Plainly, there is a huge effect of
level of processing, with deeper processing (i.e., more attention to meaning)
leading to better memory. In addition, Craik and Tulving (and many other
researchers) have confirmed the Hyde and Jenkins finding that the intention
FIGURE 6.12
Activity during first exposure
(Craik & Tulving, 1975)
24
25
12
6
10
5
U
pp
or er
lo ca
w se
er
?
es
th
rn
Le
a
ow
15
0
e!
?
nt
sa
pl
ea
le
H
ou
nt
t
C
Fi
n
d
he
th
e
tte
“e
rs
.
.”
0
20
B
rh D
ym oe
e… s it
se F ?
nt it
en in
ce th
… is
?
18
Percent of words recalled
Number of words recalled
7.What is the difference
between maintenance
rehearsal and relational (or elaborative)
rehearsal?
8.What does it mean to
say, “It doesn’t matter
if you intend to memo­
rize; all that matters for
memory is how exactly
you engage the material
you encounter”?
9.What is deep processing, and what
impact does it have on
memory?
THE IMPACT OF DEEPER PROCESSING
Activity during first exposure
(Hyde & Jenkins, 1969)
A
TEST YOURSELF
The two sets of results shown here derive from studies described in the text, but they are part of an avalanche
of data confirming the broad pattern: Shallow processing leads to poor memory. Deeper processing (paying
attention to meaning) leads to much better memory. And what matters seems to be the level of engagement;
the specific intention to learn (because participants know their memory will be tested later on) contributes little.
Entering Long-Term Storage: The Need for Engagement
•
219
COGNITION
outside the lab
Gender Differences?
Most of this book focuses on principles that apply
memory—but only when remembering the faces
to all people—young or old, sociable or shy, smart
of other women. All these differences probably
or slow. But, of course, people differ in many ways,
reflect the “attention priorities” that Western cul-
leading us to ask: Are there differences in how
ture encourages for men and women, priorities
people remember? As one aspect of this issue, do
that derive from the conventional roles assumed
men and women differ in their memories?
(for better or worse) for each gender.
Let’s emphasize at the start that there’s no
Some results also suggest that women may
overall difference between the genders in mem-
have better memory for day-to-day events,
ory accuracy, or quantity of information retained,
especially emotional events; but this, too, might
or susceptibility to outside influences that might
be a difference in attention rather than a true
pull memory off track. (See Chapter 8 for more
difference in memory. Women are, in Western
on the influences on memory.) If we take a closer
culture, encouraged to pay attention to social
look, though, we do find some differences (e.g.,
dynamics and in many settings are encouraged to
Herlitz & Rehnman, 2014)—with some studies
be more emotionally responsive and more emo-
suggesting an advantage for women in remem-
tionally sensitive than men. It may be these points
bering verbal materials, and other studies sug-
that color how women pay attention to and think
gesting an advantage for men in remembering
about an event—and ultimately how they remem-
spatial arrangement.
ber the event.
Other differences in what the genders remem-
It seems likely, then, that most of these dif-
ber are the consequence of cultural factors. Bear
ferences (none of them profound) are a direct
in mind that people tend to remember what they
reflection of cultural bias. Even so, the differences
paid attention to, and don’t remember things they
do underscore some important messages. First,
didn’t attend. From this base, it’s not surprising
with regard to their cognition, men and women
that after viewing an event, women are more likely
are much more similar to each other than they are
than men to recall the clothing people were wear-
different. Second, it’s crucial to bear in mind that
ing or their jewelry. Men, in contrast, are more likely
what you remember now is dependent on what
than women to recall the people’s body shapes.
you paid attention to earlier. Therefore, if people
(Also see Chapter 5, pp. 170–173.) There is even
differ in what they focus on, they’ll remember
some indication that women may have better face
different things later on.
to learn adds little. That is, memory performance is roughly the same in
conditions in which participants do shallow processing with an intention to
memorize, and in conditions in which they do shallow processing without
this intention. Likewise, the outcome is the same whether people do deep
processing with the intention to memorize or without. In study after study,
what matters is how people approach the material they’re seeing or hearing.
220 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
It’s that approach — that manner of engagement — that determines whether
memory will be excellent or poor later on. The intention to learn seems, by
itself, not to matter.
The Role of Meaning and
Memory Connections
The message so far seems clear: If you want to remember the sentences you’re
reading in this text, or the materials you’re learning in the training sessions
at your job, you should pay attention to what these materials mean. That is,
you should try to do deep processing. And if you do deep processing, it won’t
matter if you’re trying hard to memorize the materials (intentional learning)
or merely paying attention to the meaning because you find the material
interesting, with no plan for memorizing (incidental learning).
But what lies behind these effects? Why does attention to meaning lead to
good recall? Let’s start with a broad proposal; we’ll then fill in the evidence
for this proposal.
Connections Promote Retrieval
Perhaps surprisingly, the benefits of deep processing may not lie in the learning process itself. Instead, deep processing may influence subsequent events.
More precisely, attention to meaning may help you by facilitating retrieval
of the memory later on. To understand this point, consider what happens
whenever a library acquires a new book. On its way into the collection, the
new book must be catalogued and shelved appropriately. These steps happen when the book arrives, but the cataloguing doesn’t literally influence
the arrival of the book into the building. The moment the book is delivered,
it’s physically in the library, catalogued or not, and the book doesn’t become
“more firmly” or “more strongly” in the library because of the cataloguing.
Even so, the cataloguing is crucial. If the book were merely tossed on a
random shelf somewhere, with no entry in the catalogue, users might never
be able to find it. Without a catalogue entry, users of the library might not
even realize that the book was in the building. Notice, then, that cataloguing happens at the time of arrival, but the benefit of cataloguing isn’t for
the arrival itself. (If the librarians all went on strike, so that no books were
being catalogued, books would continue to arrive, magazines would still
be delivered, and so on. Again: The arrival doesn’t depend on cataloguing.)
Instead, the benefit of cataloguing is for events that happen after the book’s
arrival — cataloguing makes it possible (and maybe makes it easy) to find the
book later on.
The same is true for the vast library that is your memory (cf. Miller
& Springer, 1973). The task of learning is not merely a matter of placing
information into long-term storage. Learning also needs to establish some
appropriate indexing; it must pave a path to the newly acquired information,
so that this information can be retrieved at some future point. Thus, one of
The Role of Meaning and Memory Connections
•
221
WHY DO MEMORY CONNECTIONS HELP?
When books arrive in a library, the librarians must catalogue them. This doesn’t facilitate the “entry” of books into the
library, because the books are in the building whether they’re catalogued or not. But cataloguing makes the books much
easier to find later on. Memory connections may serve the same function: The connections don’t “bring” material into
memory, but they do make the material “findable” in long-term storage later.
the main chores of memory acquisition is to lay the groundwork for memory
retrieval.
But what is it that facilitates memory retrieval? There are, in fact, several
ways to search through memory, but a great deal depends on memory connections. Connections allow one memory to trigger another, and then that
memory to trigger another, so that you’re “led,” connection by connection,
to the sought-after information. In some cases, the connections link one of
the items you’re trying to remember to some of the other items; if so, finding
the first will lead you to the others. In other settings, the connections might
link some aspect of the context-of-learning to the target information, so
that when you think again about the context (“I recognize this room — this
is where I was last week”), you’ll be led to other ideas (“Oh, yeah, I read
the funny story in this room”). In all cases, though, this triggering will happen only if the relevant connections are in place — and establishing those
connections is a large part of what happens during learning.
This line of reasoning has many implications, and we can use those implications as a basis for testing whether this proposal is correct. But right at the
start, it should be clear why, according to this account, deep processing (i.e.,
attention to meaning) promotes memory. The key is that attention to meaning
involves thinking about relationships: “What words are related in meaning to
the word I’m now considering? What words have contrasting meaning? What
is the relationship between the start of this story and the way the story turned
out?” Points like these are likely to be prominent when you’re thinking about
what some word (or sentence or event) means, and these points will help you
to find (or, perhaps, to create) connections among your various ideas. It’s these
connections, we’re proposing, that really matter for memory.
222 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
Elaborate Encoding Promotes Retrieval
Notice, though, that on this account, attention to meaning is not the only
way to improve memory. Other strategies should also be helpful, provided
that they help you to establish memory connections. As an example, consider
another classic study by Craik and Tulving (1975). Participants were shown
a word and then shown a sentence with one word left out. Their task was to
decide whether the word fit into the sentence. For example, they might see
the word “chicken” and then the sentence “She cooked the _________.” The
appropriate response would be yes, because the word does fit in this sentence.
After a series of these trials, there was a surprise memory test, with participants asked to remember all the words they had seen.
But there was an additional element in this experiment. Some of the sentences shown to participants were simple, while others were more elaborate.
For example, a more complex sentence might be: “The great bird swooped
down and carried off the struggling _________.” Sentences like this one produced a large memory benefit — words were much more likely to be remembered if they appeared with these rich, elaborate sentences than if they had
appeared in the simpler sentences (see Figure 6.13).
Apparently, then, deep and elaborate processing leads to better recall than
deep processing on its own. Why? The answer hinges on memory connections. Maybe the “great bird swooped” sentence calls to mind a barnyard
scene with the hawk carrying away a chicken. Or maybe it calls to mind
FIGURE 6.13
DEEP AND ELABORATE ENCODING
Percent of words recalled
60
50
40
30
20
10
0
Simple
Complex
Sentence type
Deep processing (paying attention to meaning) promotes memory, but it isn’t
the only factor that has this benefit. More elaborate processing (e.g., by thinking about the word in the context of a complex sentence, rich with relationships) also has a powerful effect on memory.
(after craik & tulving, 1975)
The Role of Meaning and Memory Connections
•
223
TEST YOURSELF
10.What does it mean
to say, “The creation
of memory connections often occurs at
the time of learning,
but the main benefit
of those connections
comes later, at the
time of memory
retrieval”?
11.In what ways is deep
and elaborate processing superior to
deep processing on
its own?
thoughts about predator-prey relationships. One way or another, the richness of the sentence offers the potential for many connections as it calls other
thoughts to mind, each of which can be connected to the target sentence.
These connections, in turn, provide potential retrieval paths — paths that can,
in effect, guide your thoughts toward the content to be remembered. All of
this seems less likely for the simpler sentences, which will evoke fewer connections and so establish a narrower set of retrieval paths. Consequently,
words associated with these sentences are less likely to be recalled later on.
Organizing and Memorizing
Sometimes, we’ve said, memory connections link the to-be-remembered
material to other information already in memory. In other cases, the connections link one aspect of the to-be-remembered material to another aspect of
the same material. Such a connection ensures that if any part of the material
is recalled, then all will be recalled.
In all settings, though, the connections are important, and that leads us to
ask how people go about discovering (or creating) these connections. More
than 70 years ago, a psychologist named George Katona argued that the key
lies in organization (Katona, 1940). Katona’s argument was that the processes of organization and memorization are inseparable: You memorize well
when you discover the order within the material. Conversely, if you find (or
impose) an organization on the material, you will easily remember it. These
suggestions are fully compatible with the conception we’re developing here,
since what organization provides is memory connections.
Mnemonics
MNEMOSYNE
Strategies that are used to
improve memory are known
as mnemonic strategies, or
mnemonics. The term derives
from the name of the goddess of memory in Greek
mythology—Mnemosyne.
224 •
For thousands of years, people have longed for “better” memories and,
guided by this desire, people in the ancient world devised various techniques
to improve memory — techniques known as mnemonic strategies. In fact,
many of the mnemonics still in use date back to ancient Greece. (It’s therefore appropriate that these techniques are named in honor of Mnemosyne,
the goddess of memory in Greek mythology.)
How do mnemonics work? In general, these strategies provide some way
of organizing the to-be-remembered material. For example, one broad class of
mnemonic, often used for memorizing sequences of words, links the first letters
of the words into some meaningful structure. Thus, children rely on ROY G.
BIV to memorize the sequence of colors in the rainbow (red, orange, yellow . . .),
and they learn the lines in music’s treble clef via “Every Good Boy Deserves
Fudge” or “. . . Does Fine” (the lines indicate the musical notes E, G, B, D, and F).
Biology students use a sentence like “King Philip Crossed the Ocean to Find
Gold and Silver” (or: “. . . to Find Good Spaghetti”) to memorize the sequence of
taxonomic categories: kingdom, phylum, class, order, family, genus, and species.
Other mnemonics involve the use of mental imagery, relying on “mental pictures” to link the to-be-remembered items to one another. (We’ll have
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
much more to say about “mental pictures” in Chapter 11.) For example,
imagine a student trying to memorize a list of word pairs. For the pair eagletrain, the student might imagine the eagle winging back to its nest with a
locomotive in its beak. Classic research evidence indicates that images like
this can be enormously helpful. It’s important, though, that the images show
the objects in some sort of relationship or interaction — again highlighting
the role of organization. It doesn’t help just to form a picture of an eagle
and a train sitting side-by-side (Wollen, Weber, & Lowry, 1972; for another
example of a mnemonic, see Figure 6.14).
A different type of mnemonic provides an external “skeleton” for the tobe-remembered materials, and mental imagery can be useful here, too. Imagine that you want to remember a list of largely unrelated items — perhaps
FIGURE 6.14
MNEMONIC STRATEGIES
Yukon
British
Columbia
Alberta
Nunavit
New
fou
nd
lan
da
Manitoba
Quebec
Onatrio
nd
La
br
a
r
do
Sas
katc
hew
an
Northwest
Territories
Prince
Edward
Island
Nova
Scotia
New Brunswick
With a bit of creativity, you can make up mnemonics for memorizing all sorts of things. For example, can you
name all ten of the Canadian provinces? Perhaps there is a great mnemonic available, but in the meantime,
this will do. It’s a complicated mnemonic but unified by the theme of the early-morning meal: “Breakfast
Cooks Always Sell More Omelets. Quiche Never Bought; Never Sold. Perhaps Eggs In New Forms?” (You’re
on your own for remembering the three northern territories.)
Organizing and Memorizing
•
225
the entries on your shopping list, or a list of questions you want to ask your
adviser. For this purpose, you might rely on one of the so-called peg-word
systems. These systems begin with a well-organized structure, such as this one:
One is a bun.
Two is a shoe.
Three is a tree.
Four is a door.
Five is a hive.
Six are sticks.
Seven is heaven.
Eight is a gate.
Nine is a line.
Ten is a hen.
This rhyme provides ten “peg words” (“bun,” “shoe,” etc.), and in memorizing something you can “hang” the materials to be remembered on these
“pegs.” Let’s imagine that you want to remember the list of topics you need
to discuss with your adviser. If you want to discuss your unhappiness with
chemistry class, you might form an association between chemistry and the
first peg, “bun.” You might picture a hamburger bun floating in an Erlenmeyer
flask. If you also want to discuss your plans for after graduation, you might
form an association between some aspect of those plans and the next peg,
“shoe.” (You could think about how you plan to pay your way after college
by selling shoes.) Then, when meeting with your adviser, all you have to do is
think through that silly rhyme again. When you think of “one is a bun,” it’s
highly likely that the image of the flask (and therefore of chemistry lab) will
come to mind. With “two is a shoe,” you’ll be reminded of your job plans.
And so on.
Hundreds of variations on these techniques — the first-letter mnemonics,
visualization strategies, peg-word systems — are available. Some variations are
taught in self-help books (you’ve probably seen the ads — ”How to Improve Your
Memory!”); some are taught as part of corporate management training. But all
the variations use the same basic scheme. To remember a list with no apparent
organization, you impose an organization on it by using a tightly organized skele­
ton or scaffold. And, crucially, these systems all work. They help you remember
individual items, and they also help you remember those items in a specific
sequence. Figure 6.15 shows some of the data from one early study; many other
studies confirm this pattern (e.g., Bower, 1970, 1972; Bower & Reitman, 1972;
Christen & Bjork, 1976; Higbee, 1977; Roediger, 1980; Ross & Lawrence, 1968;
Yates, 1966). All of this strengthens our central claim: Mnemonics work because
they impose an organization on the materials you’re trying to memorize. And,
consistently and powerfully, organizing improves recall.
Given the power of mnemonics, students are well advised to use these
strategies in their studies. In fact, for many topics there are online databases
226 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
FIGURE 6.15
THE POWER OF MNEMONICS
Number of items recalled in proper
sequence, after 24-hour delay
5
4
3
2
1
te
im rac
ag tiv
er e
y
In
Pe
g
sy -wo
st rd
em
re Ve
he rb
ar al
sa
l
Is
o
im lat
ag ed
es
0
Type of Learning
Mnemonics can be enormously effective. In this study, students who had
relied on peg words or interactive imagery vastly outperformed students
who’d used other memorizing strategies.
(after roediger, 1980)
containing thousands of useful mnemonics — helping medical students to
memorize symptom lists, chemistry students to memorize the periodic table,
neuroscientists to remember the brain’s anatomy, and more.
Bear in mind, though, that there’s a downside to the use of mnemonics
in educational settings. When using a mnemonic, you typically focus on just
one aspect of the material you’re trying to memorize — for example, just
the first letter of the word to be remembered — and so you may cut short
your effort toward understanding this material, and likewise your effort
toward finding multiple connections between the material and other things
you know.
To put this point differently, mnemonic use involves a trade-off. If you
focus on just one or two memory connections, you’ll spend little time thinking about other possible connections, including those that might help you
understand the material. This trade-off will be fine if you don’t care very much
about the meaning of the material. (Do you care why, in taxonomy, “order”
is a subset of “class,” rather than the other way around?) But the trade-off
is troubling if you’re trying to memorize material that is meaningful. In this
case, you’d be better served by a memory strategy that seeks out multiple connections between the material you’re trying to learn and things you already
Organizing and Memorizing
•
227
know. This effort toward multiple links will help you in two ways. First, it
will foster your understanding of the material to be remembered, and so will
lead to better, richer, deeper learning. Second, it will help you retrieve this
information later. We’ve already suggested that memory connections serve as
retrieval paths, and the more paths there are, the easier it will be to find the
target material later.
For these reasons, mnemonic use may not be the best approach in many
situations. Still, the fact remains that mnemonics are immensely useful in
some settings (What were those rainbow colors?), and this confirms our
initial point: Organization promotes memory.
Understanding and Memorizing
So far, we’ve said a lot about how people memorize simple stimulus
materials — lists of randomly selected words, or colors that have to be learned
in the right sequence. In our day-to-day lives, however, we typically want to
remember more meaningful, more complicated, material. We want to remember the episodes we experience, the details of rich scenes we’ve observed,
or the many-step arguments we’ve read in a book. Do the same memory
principles apply to these cases?
The answer is clearly yes (although we’ll have more to say about this
issue in Chapter 8). In other words, your memory for events, or pictures,
or complex bodies of knowledge is enormously dependent on your being
able to organize the material to be remembered. With these more complicated materials, though, we’ve suggested that your best bet for organization
isn’t some arbitrary skeleton like those used in mnemonics. Instead, the best
organization of these complex materials is generally dependent on understanding. That is, you remember best what you understand best.
There are many ways to show that this is true. For example, we can give
people a sentence or paragraph to read and test their comprehension by asking questions about the material. Sometime later, we can test their memory.
The results are clear: The better the participants’ understanding of a sentence
or a paragraph, if questioned immediately after viewing the material, the
greater the likelihood that they will remember the material after a delay (for
classic data on this topic, see Bransford, 1979).
Likewise, consider the material you’re learning right now in the courses
you’re taking. Will you remember this material 5 years from now, or 10, or 20?
The answer depends on how well you understand the material, and one
measure of understanding is the grade you earn in a course. With full and rich
understanding, you’re likely to earn an A; with poor understanding, your grade
is likely to be lower. This leads to a prediction: If understanding is (as we’ve
proposed) important for memory, then the higher someone’s grade in a course,
the more likely that person is to remember the course contents, even years later.
This is exactly what the data show, with A students remembering the material quite well, and C students remembering much less (Conway, Cohen, &
Stanhope, 1992).
228 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
The relationship between understanding and memory can also be demonstrated in another way: by manipulating whether people understand
the material or not. For example, in an early experiment by Bransford and
Johnson (1972, p. 722), participants read this passage:
The procedure is actually quite simple. First you arrange items into
different groups. Of course one pile may be sufficient depending on
how much there is to do. If you have to go somewhere else due to lack
of facilities that is the next step; otherwise you are pretty well set. It is
important not to overdo things. That is, it is better to do too few things
at once than too many. In the short run, this may not seem important
but complications can easily arise. A mistake can be expensive as well.
At first, the whole procedure will seem complicated. Soon, however, it
will become just another facet of life. It is difficult to foresee any end to
the necessity for this task in the immediate future, but then, one never
can tell. After the procedure is completed one arranges the materials
into different groups again. Then they can be put into their appropriate
places. Eventually they will be used once more and the whole cycle will
then have to be repeated. However, that is part of life.
You’re probably puzzled by the passage, and so are most research participants. The story is easy to understand, though, if we give it a title: “Doing
the Laundry.” In the experiment, some participants were given the title before
reading the passage; others were not. Participants in the first group easily
understood the passage and were able to remember it after a delay. Participants in the second group, reading the same words, weren’t confronting a
meaningful passage and did poorly on the memory test. (For related data, see
Bransford & Franks, 1971; Sulin & Dooling, 1974. For another example, see
Figure 6.16.)
FIGURE 6.16
MEMORY FOR DIGITS
149162536496481
Examine this series of digits for a moment, and then turn away from the
page and try to recall all 15 in their proper sequence. The chances are good
that you will fail in this task—perhaps remembering the first few and the last
few digits, but not the entire list. Things will go differently, though, if you
discover the pattern within the list. Now, you’ll easily be able to remember
the full sequence. What is the pattern? Try thinking of the series this way: 1, 4,
9, 16, 25, 36. . . . Here, as always, organizing and understanding aid memory.
Organizing and Memorizing
•
229
FIGURE 6.17
OMPREHENSION ALSO AIDS MEMORY
C
FOR PICTURES
People who perceive this picture as a pattern of meaningless blotches are
unlikely to remember the picture. People who perceive the “hidden” form do
remember the picture.
(after wiseman & neisser, 1974)
TEST YOURSELF
12.Why do mnemonics
help memory? What
are the limitations
involved in mnemonic
use?
13.What’s the evidence
that there’s a clear
linkage between how
well you understand
material when you
first meet it, and
how fully you’ll recall
that information
later on?
230 •
Similar effects can be documented with nonverbal materials. Consider
the picture shown in Figure 6.17. At first it looks like a bunch of meaningless blotches; with some study, though, you may discover a familiar object.
Wiseman and Neisser (1974) tested people’s memory for this picture. Consistent with what we’ve seen so far, their memory was good if they understood
the picture — and bad otherwise. (Also see Bower, Karlin, & Dueck, 1975;
Mandler & Ritchey, 1977; Rubin & Kontis, 1983.)
The Study of Memory Acquisition
This chapter has largely been about memory acquisition. How do we
acquire new memories? How is new information, new knowledge, established in long-term memory? In more pragmatic terms, what is the best,
most effective way to learn? We now have answers to these questions, but
our discussion has indicated that we need to place these questions into a
broader context — with attention on the substantial contribution from the
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
memorizer, and also a consideration of the interconnections among acquisition, retrieval, and storage.
The Contribution of the Memorizer
Over and over, we’ve seen that memory depends on connections among ideas,
connections fostered by the steps you take in your effort toward organizing
and understanding the materials you encounter. Hand in hand with this, it
appears that memories are not established by sheer contact with the items
you’re hoping to remember. If you’re merely exposed to the items without
giving them any thought, then subsequent recall of those items will be poor.
These points draw attention to the huge role played by the memorizer. If,
for example, we wish to predict whether this or that event will be recalled,
it isn’t enough to know that someone was exposed to the event. Instead, we
need to ask what the person was doing during the event. Did she only do
maintenance rehearsal, or did she engage the material in some other way? If
the latter, how did she think about the material? Did she pay attention to the
appearance of the words or to their meaning? If she thought about meaning,
was she able to understand the material? These considerations are crucial for
predicting the success of memory.
The contribution of the memorizer is also evident in another way. We’ve
argued that learning depends on making connections, but connections to
what? If you want to connect the to-be-remembered material to other knowledge, to other memories, then you need to have that other knowledge — you
need to have other (potentially relevant) memories that you can “hook” the
new material on to.
This point helps us understand why sports fans have an easy time learning new facts about sports, and why car mechanics can easily learn new
facts about cars, and why memory experts easily memorize new information
about memory. In each situation, the person enters the learning situation with
a considerable advantage — a rich framework that the new materials can be
woven into. But, conversely, if someone enters a learning situation with little
relevant background, then there’s no framework, nothing to connect to, and
learning will be more difficult. Plainly, then, if we want to predict someone’s success in memorizing, we need to consider what other knowledge the
individual brings into the situation.
The Links among Acquisition, Retrieval, and Storage
These points lead us to another important theme. The emphasis in this chapter has been on memory acquisition, but we’ve now seen that claims about
acquisition cannot be separated from claims about storage and retrieval.
For example, why is memory acquisition improved by organization? We’ve
suggested that organization provides retrieval paths, making the memories
“findable” later on, and this is a claim about retrieval. Therefore, our claims
about acquisition are intertwined with claims about retrieval.
TEST YOURSELF
14.Explain why memorizing involves a contribution from the memorizer, both in terms of
what the memorizer
does while memorizing, and also in terms
of what the memorizer knows prior to the
memorizing.
The Study of Memory Acquisition
•
231
Likewise, we just noted that your ability to learn new material depends,
in part, on your having a framework of prior knowledge to which the new
materials can be tied. In this way, claims about memory acquisition need to
be coordinated with claims about the nature of what is already in storage.
These interactions among acquisition, knowledge, and retrieval are crucial
for our theorizing. But the interactions also have important implications for
learning, for forgetting, and for memory accuracy. The next two chapters
explore some of those implications.
COGNITIVE PSYCHOLOGY AND EDUCATION
how should i study?
Throughout your life, you encounter information that you hope to remember
later — whether you’re a student taking courses or an employee in training
for a new job. In these and many other settings, what helpful lessons can you
draw from memory research?
For a start, bear in mind that the intention to memorize, on its own, has
no effect. Therefore, you don’t need any special “memorizing steps.” Instead,
you should focus on making sure you understand the material, because if you
do, you’re likely to remember it.
As a specific strategy, it’s useful to spend a moment after a class, or after
you’ve done a reading assignment, to quiz yourself about what you’ve just
learned. You might ask questions like these: “What are the new ideas here?”;
“Do these new ideas fit with other things I know?”; “Do I know what evidence or arguments support the claims here?” Answering questions like these
will help you find meaningful connections within the material you’re learning, and between this material and other information already in your memory. In the same spirit, it’s often useful to rephrase material you encounter,
putting it into your own words. Doing this will force you to think about what
the words mean — again, a good thing for memory.
Surveys suggest, however, that most students rely on study strategies that
are much more passive than this — in fact, far too passive. Most students try
to learn materials by simply rereading the textbook or reading over their
notes several times. The problem with these strategies should be obvious:
As the chapter explains, memories are produced by active engagement with
materials, not by passive exposure.
As a related point, it’s often useful to study with a friend — so that he
or she can explain topics to you, and you can do the same in return. This
step has several advantages. In explaining things, you’re forced into a more
active role. Working with a friend is also likely to enhance your understanding, because each of you can help the other to understand bits you’re having
trouble with. You’ll also benefit from hearing your friend’s perspective on the
materials. This additional perspective offers the possibility of creating new
connections among ideas, making the information easier to recall later on.
232 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
Memory will also be best if you spread your studying out across multiple occasions — using spaced learning (e.g., spreading out your learning
across several days) rather than massed learning (essentially, “cramming” all
at once). It also helps to vary your focus while studying — working on your
history assignment for a while, then shifting to math, then over to the novel
your English professor assigned, and then back to history. There are several
reasons for this, including the fact that spaced learning and a changing focus
will make it likely that you’ll bring a somewhat different perspective to the
material each time you turn to it. This new perspective will let you see connections you didn’t see before; and — again — these new connections provide
retrieval paths that can promote recall.
Spaced learning also has another advantage. With this form of learning,
some time will pass between the episodes of learning. (Imagine, for example,
that you study your sociology text for a while on Tuesday night and then
return to it on Thursday, so that two days go by between these study sessions.)
This situation allows some amount of forgetting to take place, and that’s
actually helpful because now each episode of learning will have to take a bit
more effort, a bit more thought. This stands in contrast to massed learning,
in which your second and third passes through the material may only be
separated by a few minutes. In this setting, the second and third passes may
feel easy enough so that you zoom through them, with little engagement in
the material.
Note an ironic point here: Spaced learning may be more difficult (because
of the forgetting in between sessions), but this difficulty leads to better learning overall. Researchers refer to this as “desirable difficulty” — difficulty that
may feel obnoxious when you’re slogging through the material you hope
to learn but that is nonetheless beneficial, because it leaves you with more
complete, more long-lasting memory.
What about mnemonic strategies, such as a peg-word system? These are
enormously helpful — but often at a cost. When you’re first learning something new, focusing on a mnemonic can divert your time and attention away
from efforts at understanding the material, and so you’ll end up understanding the material less well. You’ll also be left with only the one or two retrieval
paths that the mnemonic provides, not the multiple paths created by comprehension. In some circumstances these drawbacks aren’t serious — and so,
for example, mnemonics are often useful for memorizing dates, place names,
or particular bits of terminology. But for richer, more meaningful material,
mnemonics may hurt you more than they help.
Mnemonics can be more helpful, though, after you’ve understood the new
material. Imagine that you’ve thoughtfully constructed a many-step argument or a complex derivation of a mathematical formula. Now, imagine that
you hope to re-create the argument or the derivation later on — perhaps for
an oral presentation or on an exam. In this situation, you’ve already achieved
a level of mastery, and you don’t want to lose what you’ve gained. Here, a
mnemonic (like the peg-word system) might be quite helpful, allowing you to
remember the full argument or derivation in its proper sequence.
MEANINGFUL
CONNECTIONS
What sort of connections will
help you to remember? The
answer is that almost any
connection can be helpful.
Here’s a silly—but useful—
example. Students learning
about the nervous system
have to learn that efferent
fibers carry information away
from the brain and central
nervous system, while afferent fibers carry information
inward. How to keep these
terms straight? It may be
helpful to notice that efferent
fibers carry information exiting the nervous system, while
afferent fibers provide information arriving in the nervous
system. And, as a bonus, the
same connections will help
you remember that you can
have an effect on the world
(an influence outward, from
you), but that the world can
also affect you (an influence
coming inward, toward you).
Cognitive Psychology and Education
•
233
Finally, let’s emphasize that there’s more to say about these issues.
Our discussion here (like Chapter 6 itself) focuses on the “input” side of
memory — getting information into storage, so that it’s available for use later
on. There are also steps you can take that will help you to locate information in the vast warehouse of your memory, and still other steps that you can
take to avoid forgetting materials you’ve already learned. Discussion of those
steps, however, depends on materials we’ll cover in Chapters 7 and 8.
For more on this topic . . .
Brown, P. C., Roediger, H. L., & McDaniel, M. A. (2014). Make it stick: The science
of successful learning. New York, NY: Belknap Press.
McCabe, J. A., Redick, T. S., & Engle, R. W. (2016). Brain-training pessimism, but
applied memory optimism. Psychological Science in the Public Interest, 17,
187–191.
Putnam, A. L., Sungkhasettee, V. W., & Roediger, H. L. (2016). Optimizing learning in college: Tips from cognitive psychology. Perspective on Psychological
Science, 11(5), 652–660.
234 •
C H A P T E R S I X The Acquisition of Memories and the Working-Memory System
chapter review
SUMMARY
• It is convenient to think of memorizing as having separate stages. First, one acquires new information (acquisition). Next, the information remains in
storage until it is needed. Finally, the information is
retrieved. However, this separation among the stages
may be misleading. For example, in order to memorize new information, you form connections bet­
ween this information and things you already know.
In this way, the acquisition stage is intertwined with
the retrieval of information already in storage.
• Information that is currently being considered is
held in working memory; information that isn’t currently active but is nonetheless in storage is in longterm memory. The distinction between these two
forms of memory has traditionally been described
in terms of the modal model and has been examined
in many studies of the serial-position curve. The primacy portion of this curve reflects items that have
had extra opportunity to reach long-term memory;
the recency portion of this curve reflects the accurate retrieval of items currently in working memory.
• Psychologists’ conception of working memory has
evolved in important ways in the last few decades.
Crucially, psychologists no longer think of working
memory as a “storage container” or even as a “place.”
Instead, working memory is a status—and so we say
items are “in working memory” when they’re being
actively thought about. This activity is governed by
working memory’s central executive. For mere storage, the executive often relies on low-level assistants,
including the articulatory rehearsal loop and the
visuospatial buffer, which work as mental scratch pads.
The activity inherent in this overall system is reflected
in the flexible way material can be chunked in working memory. The activity is also reflected in current
measures of working memory, via operation span.
• Maintenance rehearsal serves to keep information
in working memory and requires little effort, but it
has little impact on subsequent recall. To maximize
your chances of recall, elaborative rehearsal is needed,
in which you seek connections within the material to
be remembered or connections between the material
to be remembered and things you already know.
• In many cases, elaborative processing takes the
form of attention to meaning. This attention to
meaning is called “deep processing,” in contrast to
attention to sounds or visual form, which is considered “shallow processing.” Many studies have
shown that deep processing leads to good memory
performance later on, even if the deep processing
occurred with no intention of memorizing the target
material. In fact, the intention to learn has no direct
effect on performance; what matters instead is how
someone engages or thinks about the material to be
remembered.
• Deep processing has beneficial effects by creating effective retrieval paths that can be used later
on. Retrieval paths depend on connections linking
one memory to another; each connection provides
a path potentially leading to a target memory. Mnemonic strategies rely on the same mechanism and
focus on the creation of specific memory connections, often tying the to-be-remembered material to
a frame (e.g., a strongly structured poem).
• Perhaps the best way to form memory connections
is to understand the material to be remembered. In
understanding, you form many connections within
the material to be remembered, as well as between
this material and other knowledge. With all these retrieval paths, it becomes easy to locate this material
in memory. Consistent with these suggestions, studies have shown a close correspondence between the
ability to understand some material and the ability
to recall that material later on. This pattern has been
demonstrated with stories, visual patterns, number
series, and many other stimuli.
235
KEY TERMS
acquisition (p. 197)
storage (p. 197)
retrieval (p. 197)
modal model (p. 198)
sensory memory (p. 198)
short-term memory (p. 198)
working memory (p. 199)
long-term memory (LTM) (p. 199)
free recall (p. 200)
primacy effect (p. 200)
recency effect (p. 200)
serial position (p. 201)
memory rehearsal (p. 202)
digit-span task (p. 206)
“7 plus-or-minus 2” (p. 206)
chunks (p. 206)
operation span (p. 208)
working-memory capacity (WMC) (p. 209)
working-memory system (p. 210)
articulatory rehearsal loop (p. 210)
subvocalization (p. 211)
phonological buffer (p. 211)
concurrent articulation (p. 212)
maintenance rehearsal (p. 215)
relational (or elaborative) rehearsal (p. 215)
intentional learning (p. 218)
incidental learning (p. 218)
shallow processing (p. 219)
deep processing (p. 219)
level of processing (p. 219)
retrieval paths (p. 224)
mnemonic strategies (p. 224)
peg-word systems (p. 226)
TEST YOURSELF AGAIN
1.Define the terms “acquisition,” “storage” and
“retrieval.”
2.List the four ways in which (either in the
modal model or in more recent views) working
memory is different from long-term storage.
3.How is the primacy effect usually explained?
How is the recency effect usually explained?
4.What does it mean to say that working memory
holds seven (plus-or-minus two) “chunks”?
What is a chunk?
5.What evidence suggests that operation span is
a better measure of working memory than the
more standard digit-span measure?
6.How does the rehearsal loop manage to hold
on to information with only occasional involvement by the central executive?
7.What is the difference between maintenance
rehearsal and relational (or elaborative)
rehearsal?
236
8.What does it mean to say, “It doesn’t matter
if you intend to memorize; all that matters for
memory is how exactly you engage the material you encounter”?
9.What is deep processing, and what impact does
it have on memory?
10.What does it mean to say, “The creation of
memory connections often occurs at the time
of learning, but the main benefit of those connections comes later, at the time of memory
retrieval”?
11.In what ways is deep and elaborate processing
superior to deep processing on its own?
12.Why do mnemonics help memory? What are
the limitations of mnemonic use?
13.What’s the evidence that there’s a clear linkage
between how well you understand material
when you first meet it, and how fully you’ll
recall that information later on?
14.Explain why memorizing involves a contribution from the memorizer, both in terms of what
the memorizer does while memorizing, and
also in terms of what the memorizer knows
prior to the memorizing.
THINK ABOUT IT
1.Imagine that, based on what you’ve read in this
chapter, you were asked to write a “training
pamphlet” advising students how to study more
effectively, so that they would remember what
they studied more fully and more accurately.
What would you write in the pamphlet?
E eBOOK DEMONSTRATIONS & ESSAYS
Go to http://digital.wwnorton.com/cognition7 for the online demonstrations and essays
relevant to this chapter. These can be found in the ebook.
Online Demonstrations
• D
emonstration 6.1: Primacy and Recency Effects
• D
emonstration 6.2: Chunking
• D
emonstration 6.3: The Effects of Unattended
Online Applying Cognitive Psychology and the
Law Essays
• Cognitive Psychology and the Law: The VideoRecorder View
Exposure
• D
emonstration 6.4: Depth of Processing
• Demonstration 6.5: The Articulatory Rehearsal
Loop
• Demonstration 6.6: Sound-Based Coding
COGNITION LABS
Go to http://digital.wwnorton.com/cognition7 to experience interactive online psychology
labs relevant to this chapter.
Answer: Actually, none of the images shown in Figure 6.10 depict the Apple logo. The bottommiddle image has the bite and the dimple in the right positions, but it shows the stem pointing
the wrong way. The bottom-left image shows the stem and bite correctly, but it’s missing the
dimple!
237
7
chapter
Interconnections
between Acquisition
and Retrieval
what if…
The man known as H.M. was in his mid-20s when he
had brain surgery intended to control his epilepsy.
(We first met H.M. in Chapter 1; we also mentioned him briefly in
Chapters 2 and 6.) This surgery did achieve its aim, and H.M.’s seizures
were reduced. But the surgery had an unexpected and horrible consequence: H.M. lost the ability to form new memories. If asked what
he did last week, or yesterday, or even an hour ago, H.M. had no idea. He
couldn’t recognize the faces of medical staff he’d seen day after day.
He could read and reread a book yet never realize he’d read the same
book many times before.
A related pattern of memory loss occurs among patients who suffer from Korsakoff’s syndrome. We’ll say more about this syndrome
later in the chapter, but for now let’s highlight a paradox. These
patients, like H.M., are profoundly amnesic; they’re completely unable
to recall the events of their own lives. But these patients (again, like
H.M.) have “unconscious” memories — memories that they don’t know
they have.
We reveal these unconscious memories if we test Korsakoff’s patients
indirectly. For example, if we ask them, “Which of these melodies did
you hear an hour ago?” they’ll answer randomly — confirming their
amnesia. But if we ask them, “Which of these melodies do you prefer?”
they’re likely to choose the ones that, in fact, they heard an hour ago —
indicating that they do somehow remember (and are influenced by) the
earlier experience. If we ask them, “Have you ever seen a puzzle like this
one before?” they’ll say no. But if we ask them to solve the puzzle, their
speed will be much faster the second time — even though they insist it’s
the first time they’ve seen the puzzle. Their speed will be even faster
the third time they solve the puzzle and the fourth, although again
and again they’ll claim they’re seeing the puzzle for the very first time.
Likewise, they’ll fail if we ask them, “I showed you some words a few
minutes ago; can you tell me which of those words began ‘CHE . . .’?”
But, alternatively, we can ask them, “What’s the first word that comes to
mind that starts ‘CHE . . .’?” With this question, they’re likely to respond
with the word they’d seen earlier — a word that they ostensibly could
not remember.
239
preview of chapter themes
•
earning does not simply place information in memory;
L
instead, learning prepares you to retrieve the information
in a particular way. As a result, learning that is good preparation for one sort of retrieval may be inadequate for other
sorts of retrieval.
•
ome experiences seem to produce unconscious memoS
ries. Consideration of these “implicit memory” effects will
help us understand the various ways in which memory
influences you and will also help us see where the feeling
of familiarity comes from.
•
In general, retrieval is more likely to succeed if your perspective is the same during learning and during retrieval,
just as we would expect if learning establishes retrieval
paths that help you later when you “travel” the same path
in your effort toward locating the target material.
•
inally, an examination of amnesia confirms a central
F
theme of the chapter — namely, that we cannot speak
of “good” or “bad” memory in general. Instead, we need
to evaluate memory by considering how, and for what
purposes, the memory will be used.
These observations strongly suggest that there must be different types
of memory — including a type that’s massively disrupted in these amnesic
patients, yet one that is apparently intact (also see Figure 7.1). But how
many types of memory are there? How does each one function? Is it possible
that processes or strategies that create one type of memory might be less
useful for some other type? These questions will be central in this chapter.
Number of errors in
each attempt
40
Day 1
Day 2
Day 3
30
20
10
0
1
B
FIGURE 7.1
A
240 •
10 1
10 1
10
Attempts each day
MIRROR DRAWING
(Panel A) In a mirror-drawing task, participants must draw a
precisely defined shape — they might be asked, for example,
to trace a line between the inner and outer star. The trick,
though, is that the participants can see the figure (and their
own hand) only in the mirror. (Panel B) Performance is
usually poor at first but gradually gets better. Remarkably,
the same pattern of improvement is observed with amnesic
patients, even though on each attempt they insist that
they’re performing this task for the very first time.
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
Learning as Preparation for Retrieval
Putting information into long-term memory helps you only if you can
retrieve that information later on. Otherwise, it would be like putting
money into a savings account without the option of ever making withdrawals, or writing books that could never be read. But let’s emphasize
that there are different ways to retrieve information from memory. You
can try to recall the information (“What was the name of your tenth-grade
homeroom teacher?”) or to recognize it (“Was the name perhaps Miller?”).
If you try to recall the information, a variety of cues may or may not be
available (you might be told, as a hint, that the name began with an M or
rhymes with “tiller”).
In Chapter 6, we largely ignored these variations in retrieval. We talked as
if material was well established in memory or was not, with little regard for
how the material would be retrieved from memory. There’s reason to believe,
however, that we can’t ignore these variations in retrieval, and in this chapter
we’ll examine the interaction between how a bit of information was learned
and how it is retrieved later.
Crucial Role of Retrieval Paths
In Chapter 6, we argued that when you’re learning, you’re making connections between the newly acquired material and other information already in
your memory. These connections make the new knowledge “findable” later
on. Specifically, the connections serve as retrieval paths: When you want to
locate information in memory, you travel on those paths, moving from one
memory to the next until you reach the target material.
These claims have an important implication. To see this, bear in mind
that retrieval paths — like any paths — have a starting point and an ending
point: The path leads you from Point A to Point B. That’s useful if you want
to move from A to B, but what if you’re trying to reach B from somewhere
else? What if you’re trying to reach Point B, but at the moment you happen
to be nowhere close to Point A? In that case, the path linking A and B may
not help you.
As an analogy, imagine that you’re trying to reach Chicago from somewhere to the west. For this purpose, what you need is some highway coming
in from the west. It won’t help that you’ve constructed a wonderful road
coming into Chicago from the east. That road might be valuable in other
circumstances, but it’s not the path you need to get from where you are right
now to where you’re heading.
Do retrieval paths in memory work the same way? If so, we might find
cases in which your learning is excellent preparation for one sort of retrieval
but useless for other types of retrieval — as if you’ve built a road coming
in from one direction but now need a road from another direction. Do the
research data show this pattern?
Learning as Preparation for Retrieval
•
241
Context-Dependent Learning
Consider classic studies on context-dependent learning (Eich, 1980; Overton,
1985). In one such study, Godden and Baddeley (1975) asked scuba divers to
learn various materials. Some of the divers learned the material while sitting
on dry land; others learned it while underwater, hearing the material via a
special communication set. Within each group, half of the divers were then
tested while above water, and half were tested below (see Figure 7.2).
Underwater, the world has a different look, feel, and sound, and this
context could easily influence what thoughts come to mind for the divers
in the study. Imagine, for example, that a diver is feeling cold while underwater. This context will probably lead him to think “cold-related” thoughts,
so those thoughts will be in his mind during the learning episode. In this
situation, the diver is likely to form memory connections between these
thoughts and the materials he’s trying to learn.
Let’s now imagine that this diver is back underwater at the time of the
memory test. Most likely he’ll again feel cold, which may once more lead
him to “cold-related” thoughts. These thoughts, in turn, are now connected
(we’ve proposed) to the target materials, and that gives us what we want: The
cold triggers certain thoughts, and because of the connections formed during
learning, those thoughts can trigger the target memories.
Of course, if the diver is tested for the same memory materials on land,
he might have other links, other memory connections, that will lead to the
target memories. Even so, on land the diver will be at a disadvantage because
the “cold-related” thoughts aren’t triggered — so there will be no benefit from
the memory connections that are now in place, linking those thoughts to the
sought-after memories.
Test while
FIGURE 7.2 THE DESIGN OF A
CONTEXT-DEPENDENT LEARNING
EXPERIMENT
Half of the participants (deep-sea divers) learned
the test material while underwater; half learned
while on land. Then, within each group, half
were tested while underwater; half were tested
on land. We expect a retrieval advantage if the
learning and test circumstances match. Therefore, we expect better performance in the top
left and bottom right cells.
242 •
On land
On land
Underwater
Learning
and test
circumstances
match
CHANGE of
circumstances
between learning
and test
CHANGE of
circumstances
between learning
and test
Learning
and test
circumstances
match
Learn while
Underwater
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
By this logic, we should expect that divers who learn material while
underwater will remember the material best if they’re again underwater at
the time of the test. This setting will enable them to use the connections they
established earlier. In terms of our previous analogy, they’ve built certain
highways, and we’ve put the divers into a situation in which they can use
what they’ve built. And the opposite is true for divers who learned while on
land; they should do best if tested on land. And that is exactly what the data
show (see Figure 7.3).
Similar results have been obtained in other studies, including those
designed to mimic the learning situation of a college student. In one experiment, research participants read a two-page article similar to the sorts
of readings they might encounter in their college courses. Half the participants read the article in a quiet setting; half read it in noisy circumstances.
When later given a short-answer test, those who read the article in quiet
did best if tested in quiet — 67% correct answers, compared to 54% correct if tested in a noisy environment. Those who read the article in a noisy
environment did better if tested in a noisy environment — 62% correct,
compared to 46%. (See Grant et al., 1998; also see Balch, Bowman, &
Mohler, 1992; Cann & Ross, 1989; Schab, 1990; Smith, 1985; Smith &
Vela, 2001.)
In another study, Smith, Glenberg, and Bjork (1978) reported the same
pattern if learning and testing took place in different rooms — with the rooms
varying in appearance, sounds, and scent. In this study, though, there was an
important twist: In one version of the procedure, the participants learned
materials in one room and were tested in a different room. Just before testing,
however, the participants were urged to think about the room in which they
had learned — what it looked like and how it made them feel. When tested,
these participants performed as well as those for whom there was no room
change (Smith, 1979). What matters, therefore, is not the physical context
Test environment
14
Land
Mean words recalled
12
Underwater
10
FIGURE 7.3
LEARNING
8
6
4
2
0
Studied on land
Studied underwater
CONTEXT-DEPENDENT
Scuba divers learned materials either while on land or
while underwater. Then, they were tested while on land or
underwater. Performance was best if the divers’ circumstances at the time of test were matched to those in place
during learning.
( after godden & baddeley , 1975)
Learning as Preparation for Retrieval
•
243
TEST YOURSELF
1. What does contextdependent learning
tell us about the
nature of retrieval
paths?
2. In what ways is a
retrieval path like an
“ordinary” path (e.g.,
a path or highway
leading to a particular
city)?
244 •
but the psychological context — a result that’s consistent with our account of
this effect. As a result, you can get the benefits of context-dependent learning
through a strategy of context reinstatement — re-creating the thoughts and
feelings of the learning episode even if you’re in a very different place at the
time of recall. That’s because what matters for memory retrieval is the mental
context, not the physical environment itself.
Encoding Specificity
The results we’ve been describing also illuminate a further point: what it is
that’s stored in memory. Let’s go back to the scuba-diving experiment. The
divers in this study didn’t just remember the words they’d learned; apparently, they also remembered something about the context in which the learning took place. Otherwise, the data in Figure 7.3 (and related findings) make
no sense: If the context left no trace in memory, there’d be no way for a
return to the context to influence the divers later.
Here’s one way to think about this point, still relying on our analogy. Your
memory contains both the information you were focusing on during learning, and the highways you’ve now built, leading toward that information.
These highways — the memory connections — can of course influence your
search for the target information; that’s what we’ve been emphasizing so far.
But the connections can do more: They can also change the meaning of what
is remembered, because in many settings “memory plus this set of connections” has a different meaning from “memory plus that set of connections.”
This change in meaning, in turn, can have profound consequences for how
you remember the past.
In one of the early experiments exploring this point, participants read
target words (e.g., “piano”) in one of two contexts: “The man lifted
the piano” or “The man tuned the piano.” In each case, the sentence led
the participants to think about the target word in a particular way, and it
was this thought that was encoded into memory. In other words, what was
placed in memory wasn’t just the word “piano.” Instead, what was recorded
in memory was the idea of “piano as something heavy” or “piano as musical instrument.”
This difference in memory content became clear when participants were
later asked to recall the target words. If they had earlier seen the “lifted” sentence, they were likely to recall the target word if given the cue “something
heavy.” The hint “something with a nice sound” was much less effective.
But if participants had seen the “tuned” sentence, the result reversed: Now,
the “nice sound” hint was effective, but the “heavy” hint wasn’t (Barclay,
Bransford, Franks, McCarrell, & Nitsch, 1974). In both cases, the cue was
effective only if it was congruent with what was stored in memory.
Other experiments show a similar pattern, traditionally called encoding
specificity (Tulving, 1983; also see Hunt & Ellis, 1974; Light & Carter-Sobell,
1970). This label reminds us that what you encode (i.e., place into memory) is
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
indeed specific — not just the physical stimulus as you encountered it, but the
stimulus together with its context. Then, if you later encounter the stimulus
in some other context, you ask yourself, “Does this match anything I learned
previously?” and you correctly answer no. And we emphasize that this “no”
response is indeed correct. It’s as if you had learned the word “other” and
were later asked whether you’d been shown the word “the.” In fact, “the”
does appear as part of “other” — because the letters t h e do appear within
“other.” But it’s the whole that people learn, not the parts. Therefore, if you’ve
seen “other,” it makes sense to deny that you’ve seen “the” — or, for that matter, “he” or “her” — even though all these letter combinations are contained
within “other.”
Learning a list of words works in the same way. The word “piano” was
contained in what the research participants learned, just as “the” is contained
in “other.” What was learned, however, wasn’t just this word. Instead, what
was learned was the broader, integrated experience: the word as the perceiver
understood it. Therefore, “piano as musical instrument” isn’t what participants learned if they saw the “lifted” sentence, so they were correct in asserting that this item wasn’t on the earlier list (also see Figure 7.4).
TEST YOURSELF
3. W
hat is encoding
specificity? How is it
demonstrated?
FIGURE 7.4
REMEMBERING
“RE-CREATES” AN
EARLIER EXPERIENCE
z = –8
z = 44
R
z = 16
A
C
E
B
D
F
The text argues that what goes
into your memory is a record
of the material you’ve encountered and also a record of
the connections you established
during learning. On this basis, it
makes sense that the brain areas
activated when you’re remembering a target overlap considerably with the brain areas that
were activated when you first
encountered the target. Here,
the top panels show brain activation while viewing one picture
(A) or another picture (C) or
while hearing a particular sound
(E). The bottom panels show brain
activation while remembering
the same targets. (after wheeler,
peterson, & buckner, 2000)
Encoding Specificity
•
245
The Memory Network
In Chapter 6, we introduced the idea that memory acquisition — and, more
broadly, learning — involves the creation (or strengthening) of memory
connections. In this chapter, we’ve returned to the idea of memory connections, building on the idea that these connections serve as retrieval
paths guiding you toward the information you seek. But what are these
connections? How do they work? And who (or what?) is traveling on
these “paths”?
According to many theorists, memory is best thought of as a vast network of ideas. In later chapters, we’ll consider how exactly these ideas are
represented (as pictures? as words? in some more abstract format?). For
now, let’s just think of these representations as nodes within the network,
just like the knots in a fisherman’s net. (In fact, the word “node” is derived
from the Latin word for knot, nodus.) These nodes are tied to each other
via connections we’ll call associations or associative links. Some people find
it helpful to think of the nodes as being like light bulbs that can be turned
on by incoming electricity, and to imagine the associative links as wires that
carry the electricity.
Spreading Activation
Theorists speak of a node becoming activated when it has received a strong
enough input signal. Then, once a node has been activated, it can activate other nodes: Energy will spread out from the just-activated node via
its associations, and this will activate the nodes connected to the justactivated node.
To put all of this more precisely, nodes receive activation from
their neighbors, and as more and more activation arrives at a particular
node, the activation level for that node increases. Eventually, the activation level will reach the node’s response threshold. Once this happens, we
say that the node fires. This firing has several effects, including the fact
that the node will now itself be a source of activation, sending energy
to its neighbors and activating them. In addition, firing of the node will
draw attention to that node; this is what it means to “find” a node within
the network.
Activation levels below the response threshold, so-called subthre­
shold activation, also play an important role. Activation is assumed
to accumulate, so that two subthreshold inputs may add together, in
a process of summation, and bring the node to threshold. Likewise,
if a node has been partially activated recently, it is in effect already
“warmed up,” so that even a weak input will now be sufficient to bring it
to threshold.
These claims mesh well with points we raised in Chapter 2, when we
considered how neurons communicate with one another. Neurons receive
246 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
activation from other neurons; once a neuron reaches its threshold, it fires,
sending activation to other neurons. All of this is precisely parallel to the
suggestions we’re describing here.
Our current discussion also parallels claims offered in Chapter 4, when we
described how a network of detectors might function in object recognition.
In other words, the network linking memories to each other will resemble the
networks we’ve described linking detectors to each other (e.g., Figures 4.9
and 4.10). Detectors, like memory nodes, receive their activation from other
detectors; they can accumulate activation from different inputs, and once
activated to threshold levels, they fire.
Returning to long-term storage, however, the key idea is that activation
travels from node to node via associative links. As each node becomes
activated and fires, it serves as a source for further activation, spreading
onward through the network. This process, known as spreading activation,
enables us to deal with a key question: How does one navigate through
the maze of associations? If you start a search at one node, how do you
decide where to go from there? The answer is that in most cases you don’t
“choose” at all. Instead, activation spreads out from its starting point in
all directions simultaneously, flowing through whatever connections are
in place.
Retrieval Cues
This sketch of the memory network leaves a great deal unspecified,
but even so it allows us to explain some well-established results. For
example, why do hints help you to remember? Why, for example, do
you draw a blank if asked, “What’s the capital of South Dakota?” but
then remember if given the cue “Is it perhaps a man’s name?” Here’s one
likely explanation. Mention of South Dakota will activate nodes in
memory that represent your knowledge about this state. Activation will
then spread outward from these nodes, eventually reaching nodes that
represent the capital city’s name. It’s possible, though, that there’s only a
weak connection between the south dakota nodes and the nodes representing pierre. Maybe you’re not very familiar with South Dakota,
or maybe you haven’t thought about this state’s capital for some time.
In either case, this weak connection will do a poor job of carrying the
activation, with the result that only a trickle of activation will flow into
the pierre nodes, and so these nodes won’t reach threshold and won’t
be “found.”
Things will go differently, though, if a hint is available. If you’re told,
“South Dakota’s capital is also a man’s name,” this will activate the man’s
name node. As a result, activation will spread out from this source at the
same time that activation is spreading out from the south dakota nodes.
Therefore, the nodes for pierre will now receive activation from two sources
simultaneously, and this will probably be enough to lift the nodes’ activation
The Memory Network
•
247
FIGURE 7.5
ACTIVATION OF A NODE FROM TWO SOURCES
JACOB
SOLOMON
FRED
Nodes
representing
“South Dakota”
MAN’S
NAME
ONE OF THE
STATES
IN THE
MIDWEST
CLOSE TO
CANADA
PIERRE
TRANH
LUIS
A participant is asked, “What is the capital of South Dakota?” This activates
the south dakota nodes, and activation spreads from there to all of the
associated nodes. However, it’s possible that the connection between south
dakota and pierre is weak, so pierre may not receive enough activation to reach
threshold. Things will go differently, though, if the participant is also given
the hint “South Dakota’s capital is also a man’s name.” Now, the pierre node
will receive activation from two sources: the south dakota nodes and the
man’s name node. With this double input, it’s more likely that the pierre node
will reach threshold. This is why the hint (“man’s name”) makes the memory
search easier.
to threshold levels. In this way, question-plus-hint accomplishes more than
the question by itself (see Figure 7.5).
Semantic Priming
The explanation we’ve just offered rests on a key assumption — namely, the
summation of subthreshold activation. In other words, we relied on the idea
that the insufficient activation received from one source can add to the insufficient activation received from another source. Either source of activation
on its own wouldn’t be enough, but the two can combine to activate the
target nodes.
Can we document this summation more directly? In a lexical-decision task,
research participants are shown a series of letter sequences on a computer
screen. Some of the sequences spell words; other sequences aren’t words (e.g.,
“blar, plome”). The participants’ task is to hit a “yes” button if the sequence
spells a word and a “no” button otherwise. Presumably, they perform this
task by “looking up” these letter strings in their “mental dictionary,” and they
248 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
base their response on whether or not they find the string in the dictionary.
We can therefore use the participants’ speed of response in this task as an
index of how quickly they can locate the word in their memories.
In a series of classic studies, Meyer and Schvaneveldt (1971; Meyer,
Schvaneveldt, & Ruddy, 1974) presented participants with pairs of letter
strings, and participants had to respond “yes” if both strings were words
and “no” otherwise. For example, participants would say “yes” in response
to “chair, bread” but “no” in response to “house, fime.” Also, if both strings
were words, sometimes the words were semantically related in an obvious
way (e.g., “nurse, doctor”) and sometimes they weren’t (“cake, shoe”).
Of interest was how this relationship between the words would influence
performance.
Consider a trial in which participants see a related pair, like “bread,
butter.” To choose a response, they first need to “look up” the word
“bread” in memory. This means they’ll search for, and presumably activate, the relevant node, and in this way they’ll decide that, yes, this string
is a legitimate word. Then, they’re ready for the second word. But in this
sequence, the node for bread (the first word in the pair) has just been
activated. This will, we’ve hypothesized, trigger a spread of activation
outward from this node, bringing activation to other, nearby nodes. These
nearby nodes will surely include butter, since the association between
“bread” and “butter” is a strong one. Therefore, once the bread node
(from the first word) is activated, some activation should also spread to
the butter node.
From this base, think about what happens when a participant turns her
attention to the second word in the pair. To select a response, she must
locate “butter” in memory. If she finds this word (i.e., finds the relevant
node), then she knows that this string, too, is a word, and she can hit the
“yes” button. But the process of activating the butter node has already
begun, thanks to the (subthreshold) activation this node just received from
bread. This should accelerate the process of bringing this node to threshold (since it’s already partway there), and so it will require less time to
activate. As a result, we expect quicker responses to “butter” in this
context, compared to a context in which “butter” was preceded by some
unrelated word.
Our prediction, therefore, is that trials with related words will produce
semantic priming. The term “priming” indicates that a specific prior event
(in this case, presentation of the first word in the pair) will produce a state
of readiness (and, therefore, faster responding) later on. There are various forms of priming (in Chapter 4, we discussed repetition priming). In
the procedure we’re considering here, the priming results from the fact
that the two words in the pair are related in meaning — therefore, this is
semantic priming.
The results confirm these predictions. Participants’ lexical-decision
responses were faster by almost 100 ms if the stimulus words were related
The Memory Network
•
249
SEMANTIC PRIMING
Mean response time for pair
of words (in ms)
FIGURE 7.6
1000
900
800
700
600
First word
primes second
No priming
Condition
Participants were given a lexical-decision task involving pairs of words.
In some pairs, the words were semantically related (and so the first
word in the pair primed the second); in other pairs, the words were
unrelated (and so there was no priming). Responses to the second
word were reliably faster if the word had been primed — providing clear
evidence of the importance of subthreshold activation.
( a f t e r m e y e r & s c h va n e v e l dt , 1971)
TEST YOURSELF
4. What is subthreshold activation of a
memory node? What
role does subthreshold
activation play in explaining why retrieval
hints are often helpful?
5. How does semantic
priming illustrate the
effectiveness of subthreshold activation?
(see Figure 7.6), just as we would expect on the model we’re developing.
(For other relevant studies, including some alternative conceptions of priming, see Hutchison, 2003; Lucas, 2000.)
Before moving on, though, we should mention that this process of spreading activation — with one node activating nearby nodes — is not the whole
story for memory search. As one complication, people have some degree of
control over the starting points for their memory searches, relying on the processes of reasoning (Chapter 12) and the mechanisms of executive control
(Chapters 5 and 6). In addition, evidence suggests that once the spreading
activation has begun, people have the option of “shutting down” some of this
spread if they’re convinced that the wrong nodes are being activated (e.g.,
Anderson & Bell, 2001; Johnson & Anderson, 2004). Even so, spreading
activation is a crucial mechanism. It plays a central role in retrieval, and it
helps us understand why memory connections are so important and so helpful.
Different Forms of Memory Testing
Let’s pause to review. In Chapter 6, we argued that learning involves the
creation or strengthening of connections. This is why memory is promoted
by understanding (because understanding consists, in large part, of seeing
250 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
how new material is connected to other things you know). We also proposed that these connections later serve as retrieval paths, guiding your
search through the vast warehouse that is memory. In this chapter, we’ve
explored an important implication of this idea: that (like all paths) the
paths through memory have both a starting point and an end point. Therefore, retrieval paths will be helpful only if you’re at the appropriate starting point; this, we’ve proposed, is the basis for the advantage produced
by context reinstatement. And, finally, we’ve now started to lay out what
these paths really are: connections that carry activation from one memory
to another.
This theoretical base also helps us with another issue: the impact of
different forms of memory testing. Both in the laboratory and in dayto-day life, you often try to recall information from memory. This means
that you’re presented with a retrieval cue that broadly identifies the information you seek, and then you need to come up with the information on
your own: “What was the name of that great restaurant your parents took
us to?”; “Can you remember the words to that song?”; “Where were you
last Saturday?”
In other circumstances, you draw information from your memory via
recognition. This term refers to cases in which information is presented to
you, and you must decide whether it’s the sought-after information: “Is this
the man who robbed you?”; “I’m sure I’ll recognize the street when we get
there”; “If you let me taste that wine, I’ll tell you if it’s the same one we had
last time.”
These two modes of retrieval — recall and recognition — are fundamentally different from each other. Recall requires memory search because
you have to come up with the sought-after item on your own; you need to
locate that item within memory. As a result, recall depends heavily on the
memory connections we’ve been emphasizing so far. Recognition, in contrast, often depends on a sense of familiarity. Imagine, for example, that
you’re taking a recognition test, and the fifth word on the test is “butler.”
In response to this word, you might find yourself thinking, “I don’t recall
seeing this word on the list, but this word feels really familiar, so I guess I
must have seen it recently. Therefore, it must have been on the list.” In this
case, you don’t have source memory; that is, you don’t have any recollection of the source of your current knowledge. But you do have a strong
sense of familiarity, and you’re willing to make an inference about where
that familiarity came from. In other words, you attribute the familiarity
to the earlier encounter, and thanks to this attribution you’ll probably
respond “yes” on the recognition test.
Familiarity and Source Memory
We need to be clear about our terms here, because source memory is actually a type of recall. Let’s say, for example, that you hear a song on the radio
and say, “I know I’ve heard this song before because it feels familiar and
Different Forms of Memory Testing
•
251
I remember where I heard it.” In this setting, you’re able to remember the
source of your familiarity, and that means you’re recalling when and where
you encountered the song. On this basis, we don’t need any new theory to
talk about source memory, because we can use the same theory that we’d
use for other forms of recall. Hearing the song was the retrieval cue that
launched a search through memory, a search that allowed you to identify
the setting in which you last encountered the song. That search (like any
search) was dependent on memory connections, and would be explained by
the spreading activation process that we’ve already described.
But what about familiarity? What does this sort of remembering
involve? As a start, let’s be clear that familiarity is truly distinct from source
memory. This is evident in the fact that the two types of memory are independent of each other — it’s possible for an event to be familiar without any
source memory, and it’s possible for you to have source memory without
any familiarity. This independence is evident when you’re watching a movie
and realize that one of the actors is familiar, but (sometimes with considerable frustration, and despite a lot of effort) you can’t recall where you’ve
seen that actor before. Or you’re walking down the street, see a familiar
face, and find yourself asking, “Where do I know that woman from? Does
she work at the grocery store I shop in? Is she the driver of the bus I often
take?” You’re at a loss to answer these questions; all you know is that the
face is familiar.
In cases like these, you can’t “place” the memory; you can’t identify the
episode in which the face was last encountered. But you’re certain the face is
familiar, even though you don’t know why — a clear example of familiarity
without source memory.
The inverse case is less common, but it too can be demonstrated. For
example, in Chapter 2 we discussed Capgras syndrome. Someone with this
syndrome might have detailed, accurate memories of what friends and family
members look like, and probably remembers where and when these other
people were last encountered. Even so, when these other people are in view
they seem hauntingly unfamiliar. In this setting, there is source memory without familiarity. (For further evidence — and a patient who, after surgery, has
intact source memory but disrupted familiarity — see Bowles et al., 2007;
also see Yonelinas & Jacoby, 2012.)
We can also document the difference between source memory and familiarity in another way. In many studies, (neurologically intact) participants
have been asked, during a recognition test, to make a “remember/know”
distinction. This involves pressing one button (to indicate “remember”) if
they actually recall the episode of encountering a particular item, and pressing a different button (“know”) if they don’t recall the encounter but just
have a broad feeling that the item must have been on the earlier list. With
one response, participants are indicating that they have a source memory;
with the other, they’re indicating an absence of source memory. Basically, a
participant using the “know” response is saying, “This item seems familiar,
252 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
“FAMILIAR . . . BUT WHERE DO I KNOW HIM FROM?!?”
The photos here show successful TV or film actors. The odds are good that for some of them you’ll immediately know
their faces as familiar but won’t be sure why. You know you’ve seen these actors in some movie or show, but which one?
(We provide the actors’ names at the chapter’s end.)
so I know it was on the earlier list even though I don’t remember the experience of seeing it” (Gardiner, 1988; Hicks & Marsh, 1999; Jacoby, Jones, &
Dolan, 1998).
Researchers can use fMRI scans to monitor participants’ brain activity while they’re taking these memory tests, and the scans indicate that
“remember” and “know” judgments depend on different brain areas. The
scans show heightened activity in the hippocampus when participants indicate that they “remember” a particular test item, suggesting that this brain
structure is crucial for source memory. In contrast, “know” responses
are associated with activity in a different area — the anterior parahippocampus, with the implication that this brain site is crucial for familiarity.
(See Aggleton & Brown, 2006; Diana, Yonelinas, & Ranganath, 2007;
Dobbins, Foley, Wagner, & Schacter, 2002; Eldridge, Knowlton, Furmanski,
Bookheimer, & Engel, 2000; Montaldi, Spencer, Roberts, & Mayes, 2006;
Wagner, Shannon, Kahn, & Buckner, 2005. Also see Rugg & Curran, 2007;
Rugg & Yonelinas, 2003.)
Familiarity and source memory can also be distinguished during learning. If certain brain areas (e.g., the rhinal cortex) are especially active during
learning, then the stimulus is likely to seem familiar later on. In contrast,
if other brain areas (e.g., the hippocampal region) are particularly active
during learning, there’s a high probability that the person will indicate
source memory for that stimulus when tested later (see Figure 7.7). (See, e.g.,
Davachi & Dobbins, 2008; Davachi, Mitchell, & Wagner, 2003; Ranganath
et al., 2003.)
We still need to ask, though, what’s going on in these various brain areas
to create the relevant memories. Activity in the hippocampus is probably
helping to create the memory connections we’ve been discussing all along,
and it’s these connections, we’ve suggested, that promote source memory. But
TEST YOURSELF
6.Define “recognition”
and “recall.”
7. W
hat evidence indicates that source
memory and familiarity are distinct from
each other?
Different Forms of Memory Testing
•
253
FIGURE 7.7
FAMILIARITY VERSUS SOURCE MEMORY
Subsequent recollection effects
Subsequent familiarity effects
If the rhinal cortex
was especially
activated during
encoding, then
the stimulus was
likely to seem
familiar when
viewed later on.
If the hippocampus
was especially
activated during
encoding, then later
on the participant
was likely to
recollect having
seen that stimulus.
0.0000008
0.0006
Posterior hippocampus
Rhinal cortex
0.0005
Level of activation
Level of activation
0.0000006
0.0000004
0.0000002
0.0000000
0.0003
0.0002
0.0001
–0.0000002
0.0000
1
A
0.0004
2
3
4
5
Recognition confidence
6
Source
incorrect
Source
correct
B
In this study, researchers tracked participants’ brain activity during encoding and then analyzed the data according
to what happened later, when the time came for retrieval.
( after ranganath et al ., 2003)
what about familiarity? What “record” does it leave in memory? The answer
to this question leads us to a very different sort of memory.
Implicit Memory
How can we find out if someone remembers a previous event? The obvious
path is to ask her — “How did the job interview go?”; “Have you ever seen
Casablanca?”; “Is this the book you told me about?” But at the start of this
chapter, we talked about a different approach: We can expose someone to an
event, and then later re-expose her to the same event and assess whether her
response on the second encounter is different from the first. Specifically, we
254 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
can ask whether the first encounter somehow primed the person — got her
ready — for the second exposure. If so, it would seem that the person must
retain some record of the first encounter — she must have some sort of memory.
Memory without Awareness
In a number of studies, participants have been asked to read through a list
of words, with no indication that their memories would be tested later on.
(They might be told that they’re merely checking the list for spelling errors.)
Then, sometime later, the participants are given a lexical-decision task: They
are shown a series of letter strings and, for each, must indicate (by pressing
one button or another) whether the string is a word or not. Some of the letter
strings in the lexical-decision task are duplicates of the words seen in the first
part of the experiment (i.e., they were on the list participants had checked for
spelling), enabling us to ask whether the first exposure somehow primed the
participants for the second encounter.
In these experiments, lexical decisions are quicker if the person has
recently seen the test word; that is, lexical decision shows the pattern that in
Chapter 4 we called “repetition priming” (e.g., Oliphant, 1983). Remarkably,
this priming is observed even when participants have no recollection for having encountered the stimulus words before. To demonstrate this, we can show
participants a list of words and then test them in two different ways. One test
assesses memory directly, using a standard recognition procedure: “Which of
these words were on the list I showed you earlier?” The other test is indirect
and relies on lexical decision: “Which of these letter strings form real words?”
In this procedure, the two tests will yield different results. At a sufficient delay,
the direct memory test is likely to show that the participants have completely
forgotten the words presented earlier; their recognition performance is essentially random. According to the lexical-decision results, however, the participants still remember the words — and so they show a strong priming effect.
In this situation, then, participants are influenced by a specific past experience
that they seem (consciously) not to remember at all — a pattern that some
researchers refer to as “memory without awareness.”
A different example draws on a task called word-stem completion. In
this task, participants are given three or four letters and must produce a
word with this beginning. If, for example, they’re given cla-, then “clam” or
“clatter” would be acceptable responses, and the question of interest for us
is which of these responses the participants produce. It turns out that people
are more likely to offer a specific word if they’ve encountered it recently;
once again, this priming effect is observed even if participants, when tested
directly, show no conscious memory of their recent encounter with that word
(Graf, Mandler, & Haden, 1982).
Results like these lead psychologists to distinguish two types of memory.
Explicit memories are those usually revealed by direct memory testing —
testing that urges participants to remember the past. Recall is a direct memory test; so is a standard recognition test. Implicit memories, however, are
Implicit Memory
•
255
typically revealed by indirect memory testing and are often manifested as
priming effects. In this form of testing, participants’ current behavior is
demonstrably influenced by a prior event, but they may be unaware of this.
Lexical decision, word-stem completion, and many other tasks provide
indirect means of assessing memory. (See, for example, Mulligan & Besken,
2013; for a different perspective on these data, though, see Cabeza &
Moscovitch, 2012.)
How exactly is implicit memory different from explicit memory? We’ll say
more about this question before we’re done; but first we need to say more
about how implicit memory feels from the rememberer’s point of view. This
will lead us back into our discussion of familiarity and source memory.
False Fame
In a classic research study, Jacoby, Kelley, Brown, and Jasechko (1989) presented participants with a list of names to read out loud. The participants were
told nothing about a memory test; they thought the experiment was concerned
with how they pronounced the names. Some time later, during the second step
of the procedure, the participants were shown a new list of names and asked
to rate each person on this list according to how famous each one was. The list
included some real, very famous people; some real but not-so-famous people;
and some fictitious names that the experimenters had invented. Crucially, the
fictitious names were of two types: Some had occurred on the prior (“pronunciation”) list, and some were simply new names. A comparison between those
two types will indicate how the prior familiarization (during the pronunciation
task) influenced the participants’ judgments of fame.
For some participants, the “famous” list was presented right after the
“pronunciation” list; for other participants, there was a 24-hour delay
between these two steps. To see how this delay matters, imagine that you’re
a participant in the immediate-testing condition: When you see one of the
fictitious-but-familiar names, you might decide, “This name sounds familiar,
but that’s because I just saw it on the previous list.” In this situation, you have
a feeling that the (familiar) name is distinctive, but you also know why it’s
distinctive — because you remember your earlier encounter with the name.
In other words, you have both a sense of familiarity and a source memory,
so there’s nothing here to persuade you that the name belongs to someone
famous, and you respond accordingly. But now imagine that you’re a participant in the other condition, with the 24-hour delay. Because of the delay, you
may not recall the earlier episode of seeing the name in the pronunciation
task. But the broad sense of familiarity remains anyway, so in this setting you
might say, “This name rings a bell, and I have no idea why. I guess this must
be a famous person.” And this is, in fact, the pattern of the data: When the
two lists are presented one day apart, participants are likely to rate the madeup names as being famous.
Apparently, the participants in this study noted (correctly) that some of
the names did “ring a bell” and so did trigger a certain feeling of familiarity.
256 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
The false judgments of fame, however, come from the way the participants
interpreted this feeling and what conclusions they drew from it. Basically,
participants in the 24-hour-delay condition forgot the real source of the
familiarity (appearance on a recently viewed list) and instead filled in a
bogus source (“Maybe I saw this person in a movie?”). And it’s easy to see
why they made this misattribution. After all, the experiment was described
to them as being about fame, and other names on the list were actually those
of famous people. From the participants’ point of view, therefore, it was
reasonable to infer in this setting that any name that “rings a bell” belongs
to a famous person.
We need to be clear, though, that this misattribution is possible only
because the feeling of familiarity produced by these names was relatively
vague, and therefore open to interpretation. The suggestion, then, is that
implicit memories may leave people with only a broad sense that a stimulus
is somehow distinctive — that it “rings a bell” or “strikes a chord.” What
happens after this depends on how they interpret that feeling.
Implicit Memory and the “Illusion of Truth”
How broad is this potential for misinterpreting an implicit memory?
Participants in one study heard a series of statements and had to judge
how interesting each statement was (Begg, Anas, & Farinacci, 1992). As an
example, one sentence was “The average person in Switzerland eats about
25 pounds of cheese each year.” (This is false; the average in 1992, when the
experiment was done, was closer to 18 pounds.) Another was “Henry Ford
forgot to put a reverse gear in his first automobile.” (This is true.)
After hearing these sentences, the participants were presented with some
more sentences, but now they had to judge the credibility of these sentences,
rating them on a scale from certainly true to certainly false. However, some
of the sentences in this “truth test” were repeats from the earlier presentation, and the question of interest is how sentence credibility is influenced by
sentence familiarity.
The result was a propagandist’s dream: Sentences heard before were
more likely to be accepted as true; that is, familiarity increased credibility.
(See Begg, Armour, & Kerr, 1985; Brown & Halliday, 1990; Fiedler, Walther,
Armbruster, Fay, & Naumann, 1996; Moons, Mackie, & Garcia-Marques,
2009; Unkelbach, 2007.) This effect was found even when participants were
warned in advance not to believe the sentences in the first list. In one procedure, participants were told that half of the statements had been made
by men and half by women. The women’s statements, they were told, were
always true; the men’s, always false. (Half the participants were told the
reverse.) Then, participants rated how interesting the sentences were, with
each sentence attributed to either a man or a woman: for example, “Frank
Foster says that house mice can run an average of 4 miles per hour” or “Gail
Logan says that crocodiles sleep with their eyes open.” Later, participants
were presented with more sentences and had to judge their truth, with these
Implicit Memory
•
257
new sentences including the earlier assertions about mice, crocodiles, and
so forth.
Let’s focus on the sentences initially identified as being false — in our
example, Frank’s claim about mice. If someone explicitly remembers this
sentence (“Oh yes — Frank said such and such”), then he should judge
the assertion to be false (“After all, the experimenter said that the men’s
statements were all lies”). But what about someone who lacks this explicit
memory? This person will have no conscious recall of the episode in which
he last encountered this sentence (i.e., will have no source memory), and
so he won’t know whether the assertion came from a man or a woman.
He therefore can’t use the source as a basis for judging the truthfulness of
the sentence. But he might still have an implicit memory for the sentence
left over from the earlier exposure (“Gee, that statement rings a bell”), and
this might increase his sense of the statement’s credibility (“I’m sure I’ve
heard that somewhere before; I guess it must be true”). This is exactly the
pattern of the data: Statements plainly identified as false when they were
first heard still created the so-called illusion of truth; that is, these statements were subsequently judged to be more credible than sentences never
heard before.
The relevance of this result to the political arena or to advertising should
be clear. A newspaper headline might inquire, “Is Mayor Wilson a crook?”
Or the headline might declare, “Known criminal claims Wilson is a crook!”
In either case, the assertion that Wilson is a crook would become familiar.
The Begg et al. data indicate that this familiarity will, by itself, increase
the likelihood that you’ll later believe in Wilson’s dishonesty. This will be
true even if the paper merely raised the question; it will be true even if the
allegation came from a disreputable source. Malicious innuendo does, in
fact, produce nasty effects. (For related findings, see Ecker, Lewandowsky,
Chang, & Pillai, 2014.)
Attributing Implicit Memory to
the Wrong Source
Apparently, implicit memory can influence us (and, perhaps, bias us) in the
political arena. Other evidence suggests that implicit memory can influence
us in the marketplace — and can, for example, guide our choices when we’re
shopping (e.g., Northup & Mulligan, 2013, 2014). Yet another example
involves the justice system, and it’s an example with troubling implications.
In an early study by Brown, Deffenbacher, and Sturgill (1977), research participants witnessed a staged crime. Two or three days later, they were shown
“mug shots” of individuals who supposedly had participated in the crime.
But as it turns out, the people in these photos were different from the actual
“criminals” — no mug shots were shown for the truly “guilty” individuals.
Finally, after four or five more days, the participants were shown a lineup
and asked to select the individuals seen in Step 1 — namely, the original crime
(see Figure 7.8).
258 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
FIGURE 7.8
A PHOTO LINEUP
On TV, crime victims view a live lineup, but it’s far more common in the United
States for the victim (or witness) to see a “photo lineup” like this one. Unfortunately, victims sometimes pick the wrong person, and this error is more likely
to occur if the suspect is familiar to the victim for some reason other than the
crime. The error is unlikely, though, if the face is very familiar, because in that
case the witness will have both a feeling of familiarity and an accurate source
memory. (“Number Two looks familiar, but that’s because I see him at the gym
all the time.”)
The data in this study show a pattern known as source confusion. The
participants correctly realized that one of the faces in the lineup looked familiar, but they were confused about the source of the familiarity. They falsely
believed they had seen the person’s face in the original “crime,” when, in
truth, they’d seen that face only in a subsequent photograph. In fact, the
likelihood of this error was quite high, with 29% of the participants (falsely)
selecting from the lineup an individual they had seen only in the mug shots.
(Also see Davis, Loftus, Vanous, & Cucciare, 2008; Kersten & Earles, 2017.
For examples of similar errors that interfere with real-life criminal investigations, see Garrett, 2011. For a broader discussion of eyewitness errors, see
Reisberg, 2014.)
TEST YOURSELF
8. W
hat is the difference between implicit
and explicit memory?
Which of these is said
to be “memory without awareness”?
9.What is the role of
implicit memory in
explaining the false
fame effect?
Implicit Memory
•
259
COGNITION
outside the lab
Cryptoplagiarism
In 1970, (former Beatle) George Harrison released
when, in fact, they’d been mentioned by some one
the song “My Sweet Lord.” It turned out that the
else in the initial session.
song is virtually identical to one released years
This pattern fits well with the chapter’s discus-
before that — ”He’s So Fine,” by the Chiffons — and
sion of implicit memory. The participants in this
in 1976 Harrison was found guilty of copyright
study (and others) have lost their explicit memory
infringement. (You can find both recordings on
of the earlier episode in which they encountered
YouTube, and you’ll instantly see that they’re
someone else’s ideas. Even so, an implicit memory
essentially the same song.) In his conclusion to the
remains and emerges as a priming effect. With
court proceedings, the judge wrote, “Did Harrison
certain words primed in memory, participants
deliberately use the music of ‘He’s So Fine’? I do
are more likely to produce those words when
not believe he did so deliberately. Nevertheless,
asked — with no realization that their production
it is clear that ‘My Sweet Lord’ is the very same
has been influenced by priming.
song as ‘He’s So Fine.’ . . . This is, under the law,
Likewise, imagine talking with a friend about
infringement of copyright, and is no less so even
your options for an upcoming writing assignment.
though subconsciously accomplished” (Bright
Your friend suggests a topic, but after a moment’s
Tunes Music Corp. v. Harrisongs Music, Ltd., 420 F.
thought you reject the suggestion, convinced
Supp. 177 — Dist. Court, SD New York 1976).
that the topic is too challenging. A few days later,
How can we understand the judge’s remarks?
though, you’re again trying to choose a topic, and
Can there be “subconscious” plagiarism? The
(thanks to the priming) your friend’s suggestion
answer is yes, and the pattern at issue is some-
comes to mind. As a result of the earlier conversa-
times referred to as “cryptoplagiarism” — inadvertent
tion with your friend, you’ve had some “warm-up”
copying that is entirely unwitting and uncontrolla-
in considering this topic, so your thinking now is a
ble, and usually copying that comes with the strong
bit more fluent — and you decide that the topic isn’t
sense that you’re the inventor of the idea, even
so challenging. As a result, you go forward with the
though you’ve taken the idea from someone else.
topic. You may, however, have no explicit memory
In one early study, participants sat in groups
of the initial conversation with your friend — and you
and were asked to generate words in particu-
may not realize either that the idea “came to you”
lar categories — for example, names of sports or
because of a priming effect or that the idea seemed
musical instruments (Brown & Murphy, 1989; also
workable because of the “warm-up” provided by
Marsh, Ward, & Landau, 1999). Later, the same
the earlier conversation. The outcome: You’ll pre­
participants were asked to recall the words they
sent the idea as though it’s entirely your own, giving
(and not others in the group) had generated, and
your friend none of the credit she deserves.
also to generate new entries in the same category.
We’ll never know if this is what happened with
In this setting, participants often “borrowed” oth-
George Harrison. Even so, the judge’s conclusion
ers’ contributions — sometimes (mis)remembering
in that case seems entirely plausible, and there’s
others’ words as though they had themselves pro-
no question that inadvertent, unconscious plagia-
duced them, sometimes offering words as “new”
rism is a real phenomenon.
260 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
Theoretical Treatments of Implicit Memory
One message coming from these studies is that people are often better at
remembering that something is familiar than they are at remembering why
it is familiar. This explains why it’s possible to have a sense of familiarity
without source memory (“I’ve seen her somewhere before, but I can’t figure
out where!”) and also why it’s possible to be correct in judging familiarity
but mistaken in judging source.
In addition, let’s emphasize that in many of these studies participants are
being influenced by memories they aren’t aware of. In some cases, participants
realize that a stimulus is somehow familiar, but they have no memory of the
encounter that produced the familiarity. In other cases, they don’t even have a
sense of familiarity for the target stimulus; nonetheless, they’re influenced by
their previous encounter with the stimulus. For example, experiments show
that participants often prefer a previously presented stimulus over a novel stimulus, even though they have no sense of familiarity with either stimulus. In such
cases, people have no idea that their preference is being guided by memory
(Murphy, 2001; also Montoy, Horton, Vevea, Citkowicz, & Lauber, 2017).
It does seem, then, that the phrase “memory without awareness” is appropriate, and it does make sense to describe these memories as implicit memories. But how can we explain this form of unconscious “remembering”?
Processing Fluency
Our discussion so far — here and in Chapters 4 and 5 — has laid the foundation for a proposal about implicit memory. Let’s build the argument in steps.
When a stimulus arrives in front of your eyes, it triggers certain detectors,
and these trigger other detectors, and these still others, until you recognize the
object. (“Oh, it’s my stuffed bear, Blueberry.”) We can think of this sequence
as involving a “flow” of activation that moves from detector to detector. We
could, if we wished, keep track of this flow and in this way identify the “path”
that the activation traveled through the network. Let’s refer to this path as a
processing pathway — the sequence of detectors, and the connections between
detectors, that the activation flows through in recognizing a specific stimulus.
In the same way, we’ve proposed in this chapter that remembering often
involves the activation of a node, and this node triggers other, nearby, nodes
so that they become activated; they trigger still other nodes, leading eventually to the information you seek in memory. So here, too, we can speak
of a processing pathway — the sequence of nodes, and connections between
nodes, that the activation flows through during memory retrieval.
We’ve also said the use of a processing pathway strengthens that pathway.
This is because the baseline activation level of nodes or detectors increases if
the nodes or detectors have been used frequently in the past, or if they’ve been
used recently. Likewise, connections (between detectors or nodes) grow stronger
with use. For example, by thinking about the link between, say, “Jacob” and
“Boston,” you can strengthen the connection between the corresponding nodes,
and this will help you remember that your friend Jacob comes from Boston.
Theoretical Treatments of Implicit Memory
•
261
Now, let’s put the pieces together. Use of a processing pathway strengthens
the pathway. As a result, the pathway will be a bit more efficient, a bit faster,
the next time you use it. Theorists describe this fact by saying that use of a
pathway increases the pathway’s processing fluency — that is, the speed and
ease with which the pathway will carry activation.
In many cases, this is all the theory we need to explain implicit memory
effects. Consider implicit memory’s effect on lexical decision. In this procedure, you first are shown a list of words, including the word “bubble.” Then,
we ask you to do the lexical-decision task, and we find that you’re faster
for words (like “bubble”) that had been included in the earlier list. This increase in speed provides evidence for implicit memory, and the explanation
is straightforward. When we show you “bubble” early in the experiment,
you read the word, and this involves activation flowing through the appropriate processing pathway for this word. This warms up the pathway, and
as a result the path’s functioning will be more fluent the next time you use
it. Of course, when “bubble” shows up later as part of the lexical-decision
task, it’s handled by the same (now more fluent) pathway, and so the word is
processed more rapidly — exactly the outcome that we’re trying to explain.
For other implicit-memory effects, though, we need a further assumption —
namely, that people are sensitive to the degree of processing fluency. That is,
just as people can tell whether they’ve lifted a heavy carton or a lightweight
one, or whether they’ve answered an easy question (“What’s 2 + 2?”) or a
harder one (“What’s 17 3 19?”), people also have a broad sense of when they
have perceived easily and when they have perceived only by expending more
effort. They likewise know when a sequence of thoughts was particularly fluent
and when the sequence was labored.
This fluency, however, is perceived in an odd way. For example, when
a stimulus is easy to perceive, you don’t experience something like “That
stimulus sure was easy to recognize!” Instead, you merely register a vague
sense of specialness. You feel that the stimulus “rings a bell.” No matter how
it is described, though, this sense of specialness has a simple cause — namely,
the detection of fluency, created by practice.
There’s one complication, however. What makes a stimulus feel “special”
may not be fluency itself. Instead, people seem sensitive to changes in fluency
(e.g., they notice if it’s a little harder to recognize a face this time than it
was in the past). People also seem to notice discrepancies between how easy
(or hard) it was to carry out some mental step and how easy (or hard) they
expected it to be (Wanke & Hansen, 2015; Whittlesea, 2002). In other words,
a stimulus is registered as distinctive, or “rings a bell,” when people detect a
change or a discrepancy between experience and expectations.
To see how this matters, imagine that a friend unexpectedly gets a haircut (or gets new eyeglasses, or adds or removes some facial hair). When you
see your friend, you realize immediately that something has changed, but
you’re not sure what. You’re likely to ask puzzled questions (“Are those new
glasses?”) and get a scornful answer. (“No, you’ve seen these glasses a hundred times over the last year.”) Eventually your friend tells you what the
262 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
change is — pointing out that you failed to notice that he’d shaved off his
mustache (or some such).
What’s going on here? You obviously can still recognize your friend, but
your recognition is less fluent than in the past because of the change in your
friend’s appearance, and you notice this change — but then are at a loss to
explain it (see Figure 7.9).
On all of these grounds, we need another step in our hypothesis, but it’s
a step we’ve already introduced: When a stimulus feels special (because of
a change in fluency, or a discrepancy between the fluency expected and the
fluency experienced), you often want to know why. Thus the vague feeling of
specialness (again, produced by fluency) can trigger an attribution process, as
you ask, “Why did that stimulus stand out?”
In many circumstances, you’ll answer this question correctly, and so the
specialness will be (accurately) interpreted as familiarity and attributed to
the correct source. (“That woman seems distinctive, and I know why: It’s the
woman I saw yesterday in the dentist’s office.”) Often, you make this attribution because you have the relevant source memory — and this memory guides
you in deciding why a stimulus (a face, a song, a smell) seems to stand out. In
other cases, you make a reasonable inference, perhaps guided by the context.
(“I don’t remember where I heard this joke before, but it’s the sort of joke
that Conor is always telling, so I bet it’s one of his and that’s why the joke
is familiar.”) In other situations, though, things don’t go so smoothly, and
FIGURE 7.9
CHANGES IN APPEARANCE
The text emphasizes our sensitivity to increases in fluency, but we
can also detect decreases. In viewing a picture of a well-known
actor, for example, you might notice immediately that something
is new in his appearance, but you might be unsure about what
exactly the change involves. In this setting, the change in appearance
disrupts your well-practiced steps of perceiving for an otherwise
familiar face, so the perception is less fluent than in the past. This
lack of fluency is what gives you the “something is new” feeling. But
then the attribution step fails: You can’t identify what produced this
feeling (so you end up offering various weak hypotheses, such as
“Is that a new haircut?” when, in fact, it’s the mustache and goatee that are new). This case provides the mirror image of the cases
we’ve been considering, in which familiarity leads to an increase in
fluency, so that something “rings a bell” but you can’t say why. In the
picture shown here, you probably recognize Denzel Washington, and
you probably also realize that something is “off” in the picture. Can
you figure out what it is? (We’ve actually made several changes to
Denzel’s appearance; can you spot them all?)
Theoretical Treatments of Implicit Memory
•
263
so — as we have seen — people sometimes misinterpret their own processing
fluency, falling prey to the errors and illusions we have been discussing.
The Nature of Familiarity
All of these points provide us — at last — with a proposal for what “fami­
liarity” is, and the proposal is surprisingly complex. You might think that
familiarity is simply a feeling that’s produced more or less directly when you
encounter a stimulus you’ve met before. But the research findings described
in the last few sections point toward a different proposal — namely, that
“familiarity” is more like a conclusion that you draw rather than a feeling
triggered by a stimulus. Specifically, the evidence suggests that a stimulus
will seem familiar whenever the following list of requirements is met: First,
you have encountered the stimulus before. Second, because of that prior
encounter (and the “practice” it provided), your processing of that stimulus
is now faster and more efficient; there is, in other words, an increase in processing fluency. Third, you detect that increased fluency, and this leads you
to register the stimulus as somehow distinctive or special. Fourth, you try
to figure out why the stimulus seems special, and you reach a particular
conclusion — namely, that the stimulus has this distinctive quality because it’s
a stimulus you’ve met before in some prior episode (see Figure 7.10).
FIGURE 7.10
THE CHAIN OF EVENTS LEADING TO THE SENSE OF “FAMILIARITY”
The Steps Leading to a Judgment of Familiarity
Exposure
to a
stimulus
Practice
in
perceiving
Fluency
Stimulus
registered
as “special”
Attribution
of fluency,
perhaps
attribution
to a specific
prior event
“Familiarity”
Stimulus
registered
as “special”
Attribution
of fluency,
perhaps
attribution
to a specific
prior event
“Familiarity”
The Creation of an Illusion of Familiarity
Manipulation of stimulus
presentation designed to
make perceiving easier
Fluency
In the top line, practice in perceiving leads to fluency, and if the person attributes the fluency to some specific
prior event, the stimulus will “feel familiar.” The bottom line, however, indicates that fluency can be created in other
ways: by presenting the stimulus more clearly or for a longer exposure. Once this fluency is detected, though, it
can lead to steps identical to those in the top row. In this way, an “illusion of familiarity” can be created.
264 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
Let’s be clear, though, that none of these steps happens consciously — you’re
not aware of seeking an interpretation or trying to explain why a stimulus
feels distinctive. All you experience consciously is the end product of all these
steps: the sense that a stimulus feels familiar. Moreover, this conclusion about
a stimulus isn’t one you draw capriciously; instead, you’re likely to arrive at
this conclusion and decide a stimulus is familiar only when you have supporting information. Thus, imagine that you encounter a stimulus that “rings a
bell.” As we mentioned before, you’re likely to decide the stimulus is familiar
if you also have an (explicit) source memory, so that you can recall where and
when you last encountered that stimulus. You’re also more likely to decide a
stimulus is familiar if the surrounding circumstances support it. For example,
if you’re asked, “Which of these words were on the list you saw earlier?” the
question itself gives you a cue that some of the words were recently encountered, and so you’re more likely to attribute fluency to that encounter.
The fact remains, though, that judgments like these sometimes go astray,
which is why we need this complicated theory. We’ve considered several cases
in which a stimulus is objectively familiar (you’ve seen it recently) but doesn’t
feel familiar — just as our theory predicts. In these cases, you detect the fluency
but attribute it to some other source. (“That melody is lovely” rather than “The
melody is familiar.”) In other words, you go through all of the steps shown in
the top of Figure 7.10 except for the last two: You don’t attribute the fluency to
a specific prior event, and so you don’t experience a sense of familiarity.
We can also find the opposite sort of case — in which a stimulus is not
familiar (i.e., you’ve not seen it recently) but feels familiar anyhow — and
this, too, fits with the theory. This sort of illusion of familiarity can be
produced if the processing of a completely novel stimulus is more fluent than you
expected — perhaps because (without telling you) we’ve sharpened the focus of
a computer display or presented the stimulus for a few milliseconds longer than
other stimuli you’re inspecting (Jacoby & Whitehouse, 1989; Whittlesea, 2002;
Whittlesea, Jacoby, & Girard, 1990). Cases like these can lead to the situation
shown in the bottom half of Figure 7.10. And as our theory predicts, these situations do produce an illusion: Your processing of the stimulus is unexpectedly
fluent; you seek an attribution for this fluency, and you’re fooled into thinking the stimulus is familiar — so you say you’ve seen the stimulus before, when
in fact you haven’t. This illusion is a powerful confirmation that the sense of
familiarity does rest on processes like the ones we’ve described. (For more on
fluency, see Besken & Mulligan, 2014; Griffin, Gonzalez, Koehler, & Gilovich,
2012; Hertwig, Herzog, Schooler, & Reimer, 2008; Lanska, Olds, & Westerman,
2013; Oppenheimer, 2008; Tsai & Thomas, 2011. For a glimpse of what
fluency amounts to in the nervous system, see Knowlton & Foerde, 2008.)
The Hierarchy of Memory Types
Clearly, we’re often influenced by the past without being aware of that influence. We often respond differently to familiar stimuli than we do to novel
stimuli, even if we have no subjective feeling of familiarity. On this basis, it
Theoretical Treatments of Implicit Memory
•
265
seems that our conscious recollection seriously underestimates what’s in our
memories, and research has documented many ways in which unconscious
memories influence what we do, think, and feel.
In addition, the data are telling us that there are two different kinds
of memory: one type (“explicit”) is conscious and deliberate, the other
(“implicit”) is typically unconscious and automatic. These two broad categories can be further subdivided, as shown in Figure 7.11. Explicit memories
can be subdivided into episodic memories (memory for specific events) and
semantic memory (more general knowledge). Implicit memory is often
divided into four subcategories, as shown in the figure. Our emphasis here
has been on one of the subtypes — priming — largely because of its role in
producing the feeling of familiarity. However, the other subtypes of implicit
memory are also important and can be distinguished from priming both in
terms of their functioning (i.e., they follow somewhat different rules) and in
terms of their biological underpinnings.
Some of the best evidence for these distinctions, though, comes from the
clinic, not the laboratory. In other words, we can learn a great deal about
TEST YOURSELF
10. W
hat is processing
fluency, and how
does it influence us?
11. In what sense is
familiarity more like
a conclusion that you
draw, rather than a
feeling triggered by a
stimulus?
FIGURE 7.11
HIERARCHY OF MEMORY TYPES
Memory
Explicit memory
Conscious
Episodic memory
Memory for
specific events
Implicit memory
Revealed by indirect tests
Semantic memory
General knowledge,
not tied to any
time or place
Procedural memory
Knowing how
(i.e., memory
for skills)
Priming
Changes in
perception and
belief caused by
previous experience
Perceptual learning
Recalibration of
perceptual systems
as a result of
experience
Classical conditioning
Learning about
associations
among stimuli
In our discussion, we’ve distinguished two types of memory — explicit and implicit. However, there are reasons to
believe that each of these categories must be subdivided further, as shown here. Evidence for these subdivisions
includes functional evidence (the various types of memory follow different rules) and biological evidence (the
types depend on different aspects of brain functioning).
266 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
these various types of memory by considering individuals who have suffered
different forms of brain damage. Let’s look at some of that evidence.
Amnesia
As we have already mentioned, a variety of injuries or illnesses can lead to a loss
of memory, or amnesia. Some forms of amnesia are retrograde, meaning that
they disrupt memory for things learned prior to the event that initiated the amnesia (see Figure 7.12). Retrograde amnesia is often caused by blows to the head;
the afflicted person is unable to recall events that occurred just before the blow.
Other forms of amnesia have the reverse effect, causing disruption of memory for
experiences after the onset of amnesia; these are cases of anterograde amnesia.
(Many cases of amnesia involve both retrograde and anterograde memory loss.)
Disrupted Episodic Memory, but Spared
Semantic Memory
Studies of amnesia can teach us many things. For example, do we need all
the distinctions shown in Figure 7.11? Consider the case of Clive Wearing,
whom we met in the opening to Chapter 6. (You can find more detail
about Wearing’s case in an extraordinary book by his wife — see Wearing,
2011.) Wearing’s episodic memory is massively disrupted, but his memory
for generic information, as well as his deep love for his wife, seem to be
entirely intact. Other patients show the reverse pattern — disrupted semantic memory but preserved episodic knowledge. One patient, for example,
suffered damage (from encephalitis) to the front portion of her temporal
lobes. As a consequence, she lost her memory of many common words,
important historical events, famous people, and even the fundamental traits
of animate and inanimate objects. “However, when asked about her wedding
and honeymoon, her father’s illness and death, or other specific past episodes,
FIGURE 7.12
RETROGRADE AND ANTEROGRADE AMNESIA
Moment of
brain injury
Time
Period for which
retrograde amnesia
disrupts memory
Period for which
anterograde amnesia
disrupts memory
Retrograde amnesia disrupts memory for experiences before the injury,
accident, or disease that triggered the amnesia. Anterograde amnesia disrupts memory for experiences after the injury or disease. Some patients suffer
from both retrograde and anterograde amnesia.
Amnesia
•
267
she readily produced detailed and accurate recollections” (Schacter, 1996,
p. 152; also see Cabeza & Nyberg, 2000). (For more on amnesia, see
Brown, 2002; Clark & Maguire, 2016; Kopelman & Kapur, 2001; Nadel &
Moscovitch, 2001; Riccio, Millin, & Gisquet-Verrier, 2003.)
These cases (and other evidence too; see Figure 7.13) provide the double
dissociation that demands a distinction between episodic and semantic
memory. It’s observations like these that force us to the various distinctions
shown in Figure 7.11. (For evidence, though, that episodic and semantic
memory are intertwined in important ways, see McRae & Jones, 2012.)
Anterograde Amnesia
We’ve already mentioned the patient known as H.M. His memory loss was the
result of brain surgery in 1953, and over the next 55 years (until his death in
2008) H.M. participated in a vast number of studies. Some people suggest he
FIGURE 7.13
S EMANTIC MEMORY WITHOUT EPISODIC MEMORY
Kent Cochrane — known for years as “Patient K.C.” — died in 2014. In 1981, at age 30, he skidded off the road on his
motorcycle and suffered substantial brain damage. The damage caused severe disruption of Cochrane’s episodic
memory, but it left his semantic memory intact. As a result, he could still report on the events of his life, but these
reports were entirely devoid of autobiographical quality. In other words, he could remember the bare facts of,
say, what happened at his brother’s wedding, but the memory was totally impersonal, with no recall of context
or emotion. He also knew that during his childhood his family had fled their home because a train had derailed
nearby, spilling toxic chemicals. But, again, he simply knew this as factual material — the sort of information you
might pick up from a reference book — and he had no recall of his own experiences during the event.
268 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
was the most-studied individual in the entire history of psychology (which is one
of the reasons we’ve returned to his case several times). In fact, the data gathering
continued after H.M.’s death — with careful postmortem scrutiny of his brain.
(For a review of H.M.’s case, see Corkin, 2013; Milner, 1966, 1970; also O’Kane,
Kensinger, & Corkin, 2004; Skotko et al., 2004; Skotko, Rubin, & Tupler, 2008.)
After his surgery, H.M. was still able to recall events that took place
before the surgery — and so his amnesia was largely anterograde, not retrograde. But the amnesia was severe. Episodes he had experienced after the
surgery, people he had met, stories he had heard — all seemed to leave no
lasting record, as though nothing new could get into his long-term storage.
H.M. could hold a mostly normal conversation (because his working
memory was still intact), but his deficit became instantly clear if the conversation was interrupted. If you spoke with him for a while, then left the room
and came back 3 or 4 minutes later, he seemed to have totally forgotten that
the earlier conversation ever took place. If the earlier conversation was your
first meeting with H.M., he would, after the interruption, be certain he was
now meeting you for the very first time.
A similar amnesia has been found in patients who have been longtime alcoholics. The problem isn’t the alcohol itself; the problem instead is that alcoholics tend
to have inadequate diets, getting most of their nutrition from whatever they’re
drinking. It turns out, though, that most alcoholic beverages are missing several
key nutrients, including vitamin B1 (thiamine). As a result, longtime alcoholics
are vulnerable to problems caused by thiamine deficiency, including the disorder
known as Korsakoff’s syndrome (Rao, Larkin, & Derr, 1986; Ritchie, 1985).
Patients suffering from Korsakoff’s syndrome seem similar to H.M. in many
ways. They typically have no problem remembering events that took place
before the onset of alcoholism. They can also maintain current topics in mind
as long as there’s no interruption. New information, though, if displaced from
the mind, seems to be lost forever. Korsakoff’s patients who have been in the
hospital for decades will casually mention that they arrived only a week ago; if
Hippocampus
missing
Hippocampus
intact
A Anterior
B Posterior
H.M.’S BRAIN
For many years, researchers thought that surgery had left H.M. with no hippocampus
at all. These MRI scans of his brain show that the surgery did destroy the anterior
portion of the hippocampus (the portion closer to the front of the head; Panel A) but
not the posterior portion (closer to the rear of the head; Panel B).
Amnesia
•
269
asked the name of the current president or events in the news, they unhesitatingly give answers appropriate for two or three decades earlier, whenever the
disorder began (Marslen-Wilson & Teuber, 1975; Seltzer & Benson, 1974).
Anterograde Amnesia: What Kind of
Memory Is Disrupted?
At the chapter’s beginning, we alluded to other evidence that complicates this
portrait of anterograde amnesia, and it’s evidence that brings us back to the distinction between implicit and explicit memory. As it turns out, some of this evidence has been available for a long time. In 1911, the Swiss psychologist Édouard
Claparède (1911/1951) reported the following incident. He was introduced to a
young woman suffering from Korsakoff’s amnesia, and he reached out to shake
her hand. However, Claparède had secretly positioned a pin in his own hand so
that when they clasped hands the patient received a painful pinprick. (Modern
investigators would regard this experiment as a cruel violation of a patient’s
rights, but ethical standards were much, much lower in 1911.) The next day,
Claparède returned and reached out to shake hands with the patient. Not surprisingly, she gave no indication that she recognized Claparède or remembered
anything about the prior encounter. (This confirms the diagnosis of amnesia.)
But just before their hands touched, the patient abruptly pulled back and refused
to shake hands with Claparède. He asked her why, and after some confusion the
patient said vaguely, “Sometimes pins are hidden in people’s hands.”
What was going on here? On the one side, this patient seemed to have
no memory of the prior encounter with Claparède. She certainly didn’t mention it in explaining her refusal to shake hands, and when questioned closely
about the earlier encounter, she showed no knowledge of it. But, on the other
side, she obviously remembered something about the painful pinprick she’d
gotten the previous day. We see this clearly in her behavior.
A related pattern occurs with other Korsakoff’s patients. In one of the
early demonstrations of this point, researchers used a deck of cards like those
used in popular trivia games. Each card contained a question and some possible answers, in a multiple-choice format (Schacter, Tulving, & Wang, 1981).
The experimenter showed each card to a Korsakoff’s patient, and if the
patient didn’t know the answer, he was told it. Then, outside of the patient’s
view, the card was replaced in the deck, guaranteeing that the same question
would come up again in a few minutes.
When the question did come up again, the patients in this study were likely
to get it right — and so apparently had learned the answer in the previous
encounter. Consistent with their diagnosis, though, the patients had no recollection of the learning: They were unable to explain why their answers were
correct. They didn’t say, “I know this bit of trivia because the same question
came up just five minutes ago.” Instead, patients were likely to say things like
“I read about it somewhere” or “My sister once told me about it.”
Many studies show similar results. In setting after setting, Korsakoff’s
patients are unable to recall episodes they’ve experienced; they seem to have
270 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
no explicit memory. But if they’re tested indirectly, we see clear indications
of memory — and so these patients seem to have intact implicit memories.
(See, e.g., Cohen & Squire, 1980; Graf & Schacter, 1985; Moscovitch, 1982;
Schacter, 1996; Schacter & Tulving, 1982; Squire & McKee, 1993.) In fact, in
many tests of implicit memory, amnesic patients seem indistinguishable from
ordinary individuals.
Can There Be Explicit Memory without Implicit?
We can also find patients with the reverse pattern — intact explicit memory,
but impaired implicit. One study compared a patient who had suffered brain
damage to the hippocampus but not the amygdala with a second patient who
had the opposite pattern: damage to the amygdala but not the hippocampus
(Bechara et al., 1995). These patients were exposed to a series of trials in
which a particular stimulus (a blue light) was reliably followed by a loud
boat horn, while other stimuli (green, yellow, or red lights) were not followed
by the horn. Later on, the patients were exposed to the blue light on its own,
and their bodily arousal was measured; would they show a fright reaction
in response to this stimulus? In addition, the patients were asked directly,
“Which color was followed by the horn?”
The patient with damage to the hippocampus did show a fear reaction to
the blue light — assessed via the skin conductance response (SCR), a measure
of bodily arousal. As a result, his data on this measure look just like results
for control participants (i.e., people without brain damage; see Figure 7.14).
A
3
DAMAGE TO HIPPOCAMPUS AND AMYGDALA
Fear response
2
1
0
Explicit memory
4
Total score
SCR magnitude
(µS)
FIGURE 7.14
Controls SM046 WC1606
3
2
1
0
B
Controls SM046 WC1606
Panel A shows results for a test probing implicit memory via a fear response;
Panel B shows results for a test probing explicit memory. Patient SM046 had
suffered damage to the amygdala and shows little evidence of implicit memory
(i.e., no fear response — indexed by the skin conductance response, or SCR)
but a normal level of explicit memory. Patient WC1606 had suffered damage
to the hippocampus and shows the opposite pattern: massively disrupted
explicit memory but a normal fear response.
( after bechara et al ., 1995)
Amnesia
•
271
However, when asked directly, this patient couldn’t recall which of the lights
had been associated with the boat horn.
In contrast, the patient with damage to the amygdala showed the opposite
pattern. She was able to report that just one of the lights had been associated
with the horn and that the light’s color had been blue — demonstrating fully
intact explicit memory. When presented with the blue light, however, she
showed no fear response.
Optimal Learning
Before closing this chapter, let’s put these amnesia findings into the broader
context of the chapter’s main themes. Throughout the chapter, we’ve suggested that we cannot make claims about learning or memory acquisition
without some reference to how the learning will be used later on. For example, whether it’s better to learn underwater or on land depends on where you
will be tested. Whether it’s better to learn while listening to jazz or while
sitting in a quiet room depends on the acoustic background of the memory
test environment.
These ideas are echoed in the neuropsychology data. Specifically, it would
be misleading to say that brain damage (whether from Korsakoff’s syndrome or some other source) ruins someone’s ability to create new memories.
Instead, brain damage is likely to disrupt some types of learning but not
others, and how this matters for the person depends on how the newly
learned material will be accessed. Thus, someone who suffers hippocampal
damage will probably appear normal on an indirect memory test but seem
amnesic on a direct test, while someone who suffers amygdala damage will
probably show the reverse pattern.
All these points are enormously important for our theorizing about memory, but they also have a practical implication. Right now, you are reading this material and presumably want to remember it later on. You’re also
encountering new material in other settings (perhaps in other classes you’re
taking), and surely you want to remember that as well. How should you
study all of this information if you want the best chances of retaining it for
later use?
At one level, the message from this chapter might be that the ideal form
of learning would be one that’s “in tune with” the approach to the material
that you’ll need later. If you’re going to be tested explicitly, you want to learn
the material in a way that prepares you for that form of retrieval. If you’ll
be tested underwater or while listening to music, then, again, you want to
learn the material in a way that prepares you for that context and the mental
perspective it produces. If you’ll need source memory, then you want one
type of preparation; if you’ll need familiarity, you might want a different type
of preparation.
The problem, though, is that during learning, you often don’t know how
you’ll be approaching the material later — what the retrieval environment
272 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
will be, whether you’ll need the information implicitly or explicitly, and so on.
As a result, maybe the best strategy in learning would be to use multiple
perspectives. To revisit our earlier analogy, imagine that you know at some
point in the future you’ll want to reach Chicago, but you don’t know yet
whether you’ll be approaching the city from the north, the south, or the west.
In that case, your best bet might be to build multiple highways, so that you
can reach your goal from any direction. Memory works the same way. If you
initially think about a topic in different ways and in relation to many other
ideas, then you’ll establish many paths leading to the target material — and
so you’ll be able to access that material from many different perspectives.
The practical message from this chapter, then, is that this multiperspective
approach may provide the optimal learning strategy.
TEST YOURSELF
12. Define “retrograde”
and “anterograde”
amnesia.
13. What type(s) of
memory are disrupted
in patients suffering
from Korsakoff’s
syndrome?
COGNITIVE PSYCHOLOGY AND EDUCATION
familiarity can be treacherous
Sometimes you see a picture of someone and immediately say, “Gee — she
looks familiar!” This seems like a simple and direct reaction to the picture,
but the chapter describes how complicated familiarity really is. Indeed, the
chapter makes it clear that we can’t think of familiarity just as a “feeling”
somehow triggered by a stimulus. Instead, familiarity seems more like a conclusion that you draw at the end of a many-step process. As a result of these
complexities, errors about familiarity are possible: cases in which a stimulus
feels familiar even though it’s not, or cases in which you correctly realize that
the stimulus is familiar but then make a mistake about why it’s familiar.
These points highlight the dangers, for students, of relying on familiarity. As one illustration, consider the advice that people sometimes give for
taking a multiple-choice test. They tell you, “Go with your first inclination”
or “Choose the answer that feels familiar.” In some cases these strategies will
help, because sometimes the correct answer will indeed feel familiar. But in
other cases these strategies can lead you astray, because the answer you’re
considering may seem familiar for a bad reason. What if your professor once
said, “One of the common mistakes people make is to believe . . .” and then
talked about the claim summarized in the answer you’re now considering?
Alternatively, what if the answer seems familiar because it resembles the correct answer but is, in some crucial way, different from the correct answer
(and therefore mistaken)? In either of these cases, your sense of familiarity
might lead you to a wrong answer.
Even worse, one study familiarized people with phrases like “the record
for tallest pine tree.” Because of this exposure, these people were later more
likely to accept as true a longer phrase, such as “the record for tallest pine tree
is 350 feet.” Why? Because they realized that (at least) part of the sentence
was familiar and therefore drew the reasonable inference that they must have
FAMILIARITY CAN
BE TREACHEROUS
“Option C rings a bell. . . .”
Often, in taking a multiplechoice test, students will rea­lize
they don’t know the answer
to a question but, even so,
one of the answer options
seems somehow familiar. For
reasons described in the
chapter, though, this sense
of familiarity is an unreliable
guide in choosing a response.
Cognitive Psychology and Education
•
273
encountered the entire sentence at some previous point. The danger here
should be obvious: On a multiple-choice test, part of an incorrect option
may be an exact duplicate of some phrase in your reading; if so, relying on
familiarity will get you into trouble! (And, by the way, this claim about pines
is false; the tallest pine tree — a sugar pine — is only about 273 feet tall.)
As a different concern, think back to the end-of-chapter essay for Chapter 6.
There, we noted that one of the most common study strategies used by students is to read and reread their notes, or read and reread the textbook. This
strategy turns out not to help memory very much, and other strategies are
demonstrably better. But, in addition, the rereading strategy can actually hurt
you. Thanks to the rereading, you become more and more familiar with the
materials, which makes it easy to interpret this familiarity as mastery. But
this is a mistake, and because of the mistake, familiarity can sometimes lead
students to think they’ve mastered material when they haven’t, causing them
to end their study efforts too soon.
What can you do to avoid all these dangers? You’ll do a much better job of
assessing your own mastery if, rather than relying on familiarity, you give yourself some sort of quiz (perhaps one you find in the textbook, or one that a friend
creates for you). More broadly, it’s valuable to be alert to the various complexities associated with familiarity. After all, you don’t want to ignore familiarity,
because sometimes it’s all you’ve got. If you really don’t know the answer to a
multiple-choice question but option B seems somehow familiar, then choosing B
may be your only path forward. But given the difficulties we’ve mentioned here,
it may be best to regard familiarity just as a weak clue about the past and not
as a guaranteed indicator. That attitude may encourage the sort of caution that
will allow you to use familiarity without being betrayed by it.
For more on this topic . . .
Brown, P. C., Roediger, H. L., & McDaniel, M. A. (2014). Make it stick: The science
of successful learning. New York, NY: Belknap Press.
Jacoby, L., & Hollingshead, A. (1990). Reading student essays may be hazardous
to your spelling: Effects of reading incorrectly and correctly spelled words.
Canadian Journal of Psychology, 44, 345–358.
Preston, J., & Wegner, D. M. (2007). The Eureka error: Inadvertent plagiarism
by misattributions of effort. Journal of Personality and Social Psychology,
92, 575–584.
Stark, L.-J., Perfect, T., & Newstead, S. (2008). The effects of repeated idea
elaboration on unconscious plagiarism. Memory & Cognition, 36, 65–73.
Swire, B., Ecker, U. K. H., & Lewandowsky, S. (2017). The role of familiarity in
correcting inaccurate information. Journal of Experimental Psychology:
Learning, Memory & Cognition, 43, 1948–1961.
274 •
C H A P T E R S E V E N Interconnections between Acquisition and Retrieval
chapter review
SUMMARY
• In general, the chances that someone will remember
• Activating one node does seem to prime nearby
an earlier event are greatest if the physical and mental circumstances in place during memory retrieval
match those in place during learning. This is reflected
in the phenomenon of context-dependent learning.
nodes through the process of spreading activation.
This is evident in studies of semantic priming in
lexical-decision tasks.
• A similar pattern is reflected in the phenomenon
tion for some sorts of memory tests but ineffective
for other sorts of tests. Some strategies, for example,
are effective at establishing source memory rather
than familiarity; other strategies do the reverse.
of “encoding specificity.” This term refers to the
idea that people usually learn more than the specific
material to be remembered itself; they also learn
that material within its associated context.
• All these results arise from the fact that learning
establishes connections among memories, and these
connections serve as retrieval paths. Like any path,
these lead from some starting point to some target.
To use a given path, therefore, you must return to
the appropriate starting point. In the same way, if
there is a connection between two memories, then
activating the first memory is likely to call the second to mind. But if the first memory isn’t activated,
this connection, no matter how well established,
will not help in locating the second memory — just
as a large highway approaching Chicago from
the south won’t be helpful if you’re trying to reach
Chicago from the north.
• This emphasis on memory connections fits well
with a conceptualization of memory as a vast network, with individual nodes joined to one another
via connections or associations. An individual node
becomes activated when it receives enough of an
input signal to raise its activation level to its
response threshold. Once activated, the node sends
activation out through its connections to all the
nodes connected to it.
• Hints are effective because the target node can
receive activation from two sources simultaneously —
from nodes representing the main cue or question,
and also from nodes representing the hint.
• Some learning strategies are effective as prepara-
• Different forms of learning also play a role in
producing implicit and explicit memories. Implicit
memories are those that influence you even when
you have no awareness that you’re being influenced
by a previous event. In many cases, implicit-memory
effects take the form of priming — for example, in
a lexical decision task or word-stem completion. But
implicit memories can also influence you in other
ways, producing a number of memory-based illusions.
• Implicit memory can be understood as the consequence of increased processing fluency, produced
by experience in a particular task with a particular stimulus. The fluency is sometimes detected and
registered as a sense of “specialness” attached to
a stimulus. Often, this specialness is attributed to
some cause, but this attribution can be inaccurate.
• Implicit memory is also important in understanding the pattern of symptoms in anterograde amnesia.
Amnesic patients perform badly on tests requiring
explicit memory and may not even recall events that
happened just minutes earlier. However, they often
perform at near-normal levels on tests involving
implicit memory. This disparity underscores the fact
that we cannot speak in general about good and bad
memories, good and poor learning. Instead, learning
and memory must be matched to a particular task and
a particular form of test; learning and memory that
are excellent for some tasks may be poor for others.
275
KEY TERMS
context-dependent learning (p. 242)
context reinstatement (p. 244)
encoding specificity (p. 244)
nodes (p. 246)
associations (or associative links) (p. 246)
activation level (p. 246)
response threshold (p. 246)
subthreshold activation (p. 246)
summation (p. 246)
spreading activation (p. 247)
lexical-decision task (p. 248)
semantic priming (p. 249)
recall (p. 251)
recognition (p. 251)
source memory (p. 251)
familiarity (p. 251)
attribution (p. 251)
“remember/know” distinction (p. 252)
word-stem completion (p. 255)
explicit memory (p. 255)
direct memory testing (p. 255)
implicit memory (p. 255)
indirect memory testing (p. 256)
illusion of truth (p. 258)
source confusion (p. 259)
processing pathway (p. 261)
processing fluency (p. 262)
amnesia (p. 267)
retrograde amnesia (p. 267)
anterograde amnesia (p. 267)
Korsakoff’s syndrome (p. 269)
TEST YOURSELF AGAIN
1.What does context-dependent learning tell us
about the nature of retrieval paths?
2.In what ways is a retrieval path like an
“ordinary” path (e.g., a path or highway
leading to a particular city)?
8.What is the difference between implicit and
explicit memory? Which of these is said to
be “memory without awareness”?
9.What is the role of implicit memory in
explaining the false fame effect?
3.What is encoding specificity? How is it
demonstrated?
10.What is processing fluency, and how does it
influence us?
4.What is subthreshold activation of a memory
node? What role does subthreshold activation
play in the explanation of why retrieval hints are
often helpful?
11.In what sense is familiarity more like a
conclusion that you draw, rather than a
feeling triggered by a stimulus?
5.How does semantic priming illustrate the
effectiveness of subthreshold activation?
6.Define “recognition” and “recall.”
7.What evidence indicates that source memory
and familiarity are distinct from each other?
276
12.Define “retrograde” and “anterograde”
amnesia.
13.What type(s) of memory are disrupted
in patients suffering from Korsakoff’s
syndrome?
THINK ABOUT IT
1.Some people describe the eerie sensation of
“déjà vu” — a feeling in which a place or face
seems familiar, even though they’re quite
certain they’ve never been in this place, or
seen this face, before. Can you generate a
hypothesis about the roots of déjà vu,
drawing on the material in the chapter?
E eBOOK DEMONSTRATIONS & ESSAYS
Go to http://digital.wwnorton.com/cognition7 for the online demonstrations and essays
relevant to this chapter. These can be found in the ebook.
Online Demonstrations
Online Applying Cognitive Psychology and the
Law Essays
• Demonstration 7.1: Retrieval Paths and
• Cognitive Psychology and the Law: Unconscious
Connections
Transference
• Demonstration 7.2: Encoding Specificity
• Demonstration 7.3: Spreading Activation in
Memory Search
• Demonstration 7.4: Semantic Priming
• Demonstration 7.5: Priming From Implicit
Memory
COGNITION LABS
Go to http://digital.wwnorton.com/cognition7 to experience interactive online psychology
labs relevant to this chapter.
The performers who appear on p. 253 are (left to right). Michael K. Williams, Margo Martindale, Adam Rodriguez, Judy Greer, and
Adina Porter.
277
8
chapter
Remembering
Complex Events
what if…
What were you doing on March 23, 2009? What did
you have for lunch that day? What was the weather
like? These seem like odd questions; why should you remember these
details from almost a decade ago? But imagine if you could remember
those details — and similar details for every other day in your life. In other
words, what would life be like if you had a “super memory” — so that,
essentially, you never forgot anything?
Researchers have identified a small number of people who have
hyperthymesia, also called “highly superior autobiographical recall”
(HSAM). These people seem able to remember every single day of their
lives, over a span of many years. One of these individuals claims that she
can recall every day of her life over the last four decades. “Starting on
February 5, 1980, I remember everything. That was a Tuesday.”
If asked about a randomly selected date — say, February 10,
2012 — these individuals can recall exactly where they were that day,
what time they woke up, and what shoes they wore. Their performance
is just as good if they’re asked about October 8, 2008, or December 19,
2007. And when checked, their memories turn out to be uniformly
accurate.
These individuals are remarkable in how much they remember, but
they seem quite normal in other ways. For example, their extraordinary
memory capacity hasn’t made them amazing geniuses or incredible
scholars. Even though they have an exceptional capacity for remembering their own lives, they have no advantage in remembering other
sorts of content or performing other mental tasks. This point has been
documented with careful testing, but it is also evident in the fact that
researchers weren’t aware such people existed until just a few years
ago (e.g., Parker, Cahill, & McGaugh, 2006). Apparently, these individuals, even with their incredible memory capacity, are ordinary enough
in other ways so that we didn’t spot them until recently (McGaugh &
LePort, 2014.)
Humans have been trying for centuries to improve their memories,
but it seems that a “perfect” memory may provide less of an advantage
than you might think. We’ll return to this point, and what it implies about
memory functioning, later in the chapter.
279
preview of chapter themes
•
utside the lab, you often try to remember materials that
O
are related in some way to other things you know or have
experienced. Over and over, we will see that this other
knowledge — the knowledge you bring to a situation —
helps you to remember by promoting retrieval, but it can
also promote error.
•
he memory errors produced by prior knowledge tend to be
T
quite systematic: You often recall the past as more “normal,”
more in line with your expectations, than it actually was.
•
ven acknowledging the memory errors, our overall
E
assessment of memory can be quite positive. This is
because memories are accurate most of the time, and
the errors that do occur can be understood as the byproducts of mechanisms that generally serve you well.
•
inally, we will consider three factors that play an imporF
tant role in shaping memory outside of the laboratory:
involvement with an event, emotion, and the passage of
time. These factors require some additional principles as
part of our overall theory, but they also confirm the power
of more general principles — principles hinging, for example,
on the role of memory connections.
Memory Errors, Memory Gaps
Where did you spend last summer? What country did you grow up in? Where
were you five minutes ago? These are easy questions, and you effortlessly
retrieve this information from memory the moment you need it. If we want
to understand how memory functions, therefore, we need to understand how
you locate these bits of information (and thousands of others just like them)
so readily.
But we also need to account for some other observations. Sometimes,
when you try to remember an episode, you draw a blank. On other
occasions, you recall something, but with no certainty that you’re correct:
“I think her nickname was Dink, but I’m not sure.” And sometimes,
when you do recall a past episode, it turns out that your memory is mistaken. Perhaps a few details of the event were different from the way you
recall them. Or perhaps your memory is completely wrong, misrepresenting large elements of the original episode. Worse, in some cases you can
remember entire events that never happened at all! In this chapter, we’ll
consider how, and how often, these errors arise. Let’s start with some
examples.
Memory Errors: Some Initial Examples
In 1992, an El Al cargo plane lost power in two of its engines just after
taking off from Amsterdam’s Schiphol Airport. The pilot attempted to return
the plane to the airport but couldn’t make it. A few minutes later, the plane
crashed into an 11-story apartment building in Amsterdam’s Bijlmermeer
neighborhood. The building collapsed and burst into flames; 43 people were
killed, including the plane’s entire crew.
280 •
C H A P T E R E I G H T Remembering Complex Events
Ten months later, researchers questioned 193 Dutch people about
the crash, asking them in particular, “Did you see the television film of
the moment the plane hit the apartment building?” More than half of the
participants (107 of them) reported seeing the film, even though there
was no such film. No camera had recorded the crash; no film (or any
reenactment) was shown on television. The participants seemed to be
remembering something that never took place (Crombag, Wagenaar, &
van Koppen, 1996).
In a follow-up study, investigators surveyed another 93 people about
the plane crash. These people were also asked whether they’d seen the (nonexistent) TV film, and then they were asked detailed questions about exactly
what they had seen in the film: Was the plane burning when it crashed, or did
it catch fire a moment later? In the film, did they see the plane come down
vertically with no forward speed, or did it hit the building while still moving
horizontally at a considerable speed?
Two thirds of these participants reported seeing the film, and most of them
were able to provide details about what they had seen. When asked about the
plane’s speed, for example, only 23% said that they couldn’t remember.
The others gave various responses, presumably based on their “memory” of
the (nonexistent) film.
Other studies have produced similar results. There was no video footage of
the car crash in which Princess Diana was killed, but 44% of the British participants in one study recalled seeing the footage (Ost, Vrij, Costall, & Bull,
2002). More than a third of the participants questioned about a nightclub
bombing in Bali recalled seeing a (nonexistent) video, and nearly all these
participants reported details about what they’d seen in the video (Wilson &
French, 2006).
It turns out that more persistent questioning can lead some of these people
to admit they actually don’t remember seeing the video. Even with persistent
questioning, though, many participants continue to insist that they did see
the video — and they offer additional information about exactly what they
saw in the film (e.g., Patihis & Loftus, 2015; Smeets et al., 2006). Also, in all
these studies, let’s emphasize that participants are thinking back to an emotional and much-discussed event; the researchers aren’t asking them to recall
a minor occurrence.
Is memory more accurate when the questions come after a shorter delay? In
a study by Brewer and Treyens (1981), participants were asked to wait briefly
in the experimenter’s office prior to the procedure’s start. After 35 seconds,
participants were taken out of this office and told that there actually was no
experimental procedure. Instead, the study was concerned with their memory
for the room in which they’d just been sitting.
Participants’ descriptions of the office were powerfully influenced by their
prior beliefs. Surely, most participants would expect an academic office to
contain shelves filled with books. In this particular office, though, no books
TEST YOURSELF
1. W
hat is the evidence
that in some circumstances many people
will misremember
significant events they
have experienced?
2. W
hat is the evidence
that in some circumstances people will
even misremember
recent events?
Memory Errors, Memory Gaps
•
281
FIGURE 8.1 THE
OFFICE USED IN
THE BREWER AND
TREYENS STUDY
No books were in view in this
office, but many participants,
biased by their expectations
of what should be in an
academic office, remembered
seeing books.
(after brewer & treyens, 1981)
were in view (see Figure 8.1). Even so, almost one third of the participants
(9 of 30) reported seeing books in the office. Their recall, in other words, was
governed by their expectations, not by reality.
How could this happen? How could so many Dutch participants be wrong
in their recall of the plane crash? How could intelligent, alert college students
fail to remember what they’d seen in an office just moments earlier?
Memory Errors: A Hypothesis
In Chapters 6 and 7, we emphasized the importance of memory connections
that link each bit of knowledge in your memory to other bits. Sometimes
these connections tie together similar episodes, so that a trip to the beach
ends up connected in memory to your recollection of other trips. Sometimes
the connections tie an episode to certain ideas — ideas, perhaps, that were
part of your understanding of the episode, or ideas that were triggered by
some element within the episode.
It’s not just separate episodes and ideas that are linked in this way. Even
for a single episode, the elements of the episode are stored separately from
one another and are linked by connections. In fact, the storage is “modalityspecific,” with the bits representing what you saw stored in brain areas
devoted to visual processing, the bits representing what you heard stored
in brain areas specialized for auditory processing, and so on (e.g., Nyberg,
Habib, McIntosh, & Tulving, 2000; Wheeler, Peterson, & Buckner, 2000;
also see Chapter 7, Figure 7.4, p. 245).
282 •
C H A P T E R E I G H T Remembering Complex Events
With all these connections in place — element to element, episode to episode, episode to related ideas — information ends up stored in memory in a
system that resembles a vast spider web, with each bit of information connected by many threads to other bits elsewhere in the web. This was the idea
that in Chapter 7 we described as a huge network of interconnected nodes.
However, within this network there are no boundaries keeping the elements of one episode separate from elements of other episodes. The episodes,
in other words, aren’t stored in separate “files,” each distinct from the others. What is it, therefore, that holds together the various bits within each
episode? To a large extent, it’s simply the density of connections. There are
many connections linking the various aspects of your “trip to the beach” to
one another; there are fewer connections linking this event to other events.
As we’ve discussed, these connections play a crucial role in memory retrieval. Imagine that you’re trying to recall the restaurant you ate at during
your beach trip. You’ll start by activating nodes in memory that represent
some aspect of the trip — perhaps your memory of the rainy weather. Activation will then flow outward from there, through the connections you’ve
established, and this will energize nodes representing other aspects of the trip.
The flow of activation can then continue from there, eventually reaching the
nodes you seek. In this way, the connections serve as retrieval paths, guiding
your search through memory.
Obviously, then, memory connections are a good thing; without them,
you might never locate the information you’re seeking. But the connections
can also create problems. As you add more and more links between the bits
of this episode and the bits of that episode, you’re gradually knitting these
two episodes together. As a result, you may lose track of the “boundary” between the episodes. More precisely, you’re likely to lose track of which bits
of information were contained within which event. In this way, you become
vulnerable to what we might think of as “transplant” errors, in which a bit of
information encountered in one context is transplanted into another context.
In the same way, as your memory for an episode becomes more and more
interwoven with other thoughts you’ve had about the event, it will become
difficult to keep track of which elements are linked to the episode because
they were actually part of the episode itself, and which are linked merely because they were associated with the episode in your thoughts. This, too, can
produce transplant errors, in which elements that were part of your thinking
get misremembered as if they were actually part of the original experience.
Understanding Both Helps and Hurts Memory
It seems, then, that memory connections both help and hurt recollection.
They help because the connections, serving as retrieval paths, enable you to
locate information in memory. But connections can hurt because they sometimes make it difficult to see where the remembered episode stops and other,
related knowledge begins. As a result, the connections encourage intrusion
errors — errors in which other knowledge intrudes into the remembered event.
Memory Errors: A Hypothesis
•
283
To see how these points play out, consider an early study by Owens,
Bower, and Black (1979). In this study, half of the participants read the
following passage:
Nancy arrived at the cocktail party. She looked around the room to see
who was there. She went to talk with her professor. She felt she had
to talk to him but was a little nervous about just what to say. A group
of people started to play charades. Nancy went over and had some
refreshments. The hors d’oeuvres were good, but she wasn’t interested
in talking to the rest of the people at the party. After a while she
decided she’d had enough and left the party.
Other participants read the same passage, but with a prologue that set
the stage:
Nancy woke up feeling sick again, and she wondered if she really was
pregnant. How would she tell the professor she had been seeing? And
the money was another problem.
All participants were then given a recall test in which they were asked
to remember the sentences as exactly as they could. Table 8.1 shows the
results — the participants who had read the prologue (the Theme condition)
recalled much more of the original story (i.e., they remembered the propositions actually contained within the story). This is what we should expect,
based on the claims made in Chapter 6: The prologue provided a meaningful
context for the remainder of the story, and this helped understanding. Understanding, in turn, promoted recall.
At the same time, the story’s prologue also led participants to include elements in their recall that weren’t mentioned in the original episode. In fact,
participants who had seen the prologue made four times as many intrusion
errors as did participants who hadn’t seen the prologue. For example, they
might include in their recall something like “The professor had gotten Nancy
pregnant.” This idea isn’t part of the story but is certainly implied, so will
probably be part of participants’ understanding of the story. It’s then this
understanding (including the imported element) that is remembered.
TABLE 8.1
UMBER OF PROPOSITIONS REMEMBERED
N
BY PARTICIPANTS
STUDIED PROPOSITIONS
(THOSE IN STORY)
Theme Condition
Neutral Condition
29.2
20.2
INFERRED PROPOSITIONS
(THOSE NOT IN STORY)
Theme Condition
15.2
Neutral Condition
3.7
In the Theme condition, a brief prologue set the theme for the passage that was to
(after owens et al., 1979)
be remembered.
284 •
C H A P T E R E I G H T Remembering Complex Events
The DRM Procedure
Similar effects, with memory connections both helping and hurting memory,
can be demonstrated with simple word lists. For example, in many experiments, participants have been presented with lists like this one: “bed, rest,
awake, tired, dream, wake, snooze, blanket, doze, slumber, snore, nap, peace,
yawn, drowsy.” Immediately after hearing this list, participants are asked to
recall as many of the words as they can.
As you surely noticed, the words in this list are all associated with sleep,
and the presence of this theme helps memory: The list words are easy to remember. It turns out, though, that the word “sleep” is not itself included in the
list. Nonetheless, research participants spontaneously make the connection
between the list words and this associated word, and this connection almost
always leads to a memory error. When the time comes for recall, participants
are extremely likely to recall that they heard “sleep.” In fact, they’re just as
likely to recall “sleep” as they are to recall the actual words on the list (see
Figure 8.2). When asked how confident they are in their memories, participants are just as confident in their (false) recall of “sleep” as they are in their
(correct) memory of genuine list words (Gallo, 2010; for earlier and classic
papers in this arena, see Deese, 1957; Roediger & McDermott, 1995, 2000).
This experiment (and many others like it) uses the DRM procedure, a
bit of terminology that honors the investigators who developed it (James
Deese, Henry Roediger III, and Kathleen McDermott). The procedure yields
many errors even if participants are put on their guard before the procedure
begins — that is, told about the nature of the lists and the frequency with
which they produce errors (Gallo, Roberts, & Seamon, 1997; McDermott &
100
90
Percent of recall
80
70
60
50
40
FIGURE 8.2 THE EFFECTS OF THE
DRM PARADIGM
30
20
10
0
Words from
actual list
Unrelated
words
Mistaken “recall”
of theme words
Because of the theme uniting the list, participants can remember almost 90% of the words
they encountered. However, they’re just as likely
to “recall” the list’s theme word — even though it
was not presented.
Memory Errors: A Hypothesis
•
285
Roediger, 1998). Apparently, the mechanisms leading to these errors are so
automatic that people can’t inhibit them.
Schematic Knowledge
Imagine that you go to a restaurant with a friend. This setting is familiar for
you, and you have some commonsense knowledge about what normally happens here. You’ll be seated; someone will bring menus; you’ll order, then eat;
eventually, you’ll pay and leave. Knowledge like this is often referred to with
the Greek word schema (plural: schemata). Schemata summarize the broad
pattern of what’s normal in a situation — and so your kitchen schema tells
you that a kitchen is likely to have a stove but no piano; your dentist’s office
schema tells you that there are likely to be magazines in the waiting room,
that you’ll probably get a new toothbrush when you leave, and so on.
Schemata help you in many ways. In a restaurant, for example, you’re not
puzzled when someone keeps filling your water glass or when someone else
drops by to ask, “How is everything?” Your schema tells you that these are
normal occurrences in a restaurant, and you instantly understand how they
fit into the broader framework.
Schemata also help when the time comes to recall how an event unfolded.
This is because there are often gaps in your recollection — either because you
didn’t notice certain things in the first place, or because you’ve gradually
forgotten some aspects of the experience. (We’ll say more about forgetting
later in the chapter.) In either case, you can rely on your schemata to fill in
these gaps. So, in thinking back to your dinner at Chez Pierre, you might not
remember anything about the menus. Nonetheless, you can be reasonably
sure that there were menus and that they were given to you early on and
taken away after you placed your order. On this basis, you’re likely to include
menus within your “recall” of the dinner, even if you have no memory of
seeing the menus for this particular meal. In other words, you’ll supplement
what you actually remember with a plausible reconstruction based on your
schematic knowledge. And in most cases this after-the-fact reconstruction
will be correct, since schemata do, after all, describe what happens most of
the time.
Evidence for Schematic Knowledge
Clearly, then, schematic knowledge helps you, by guiding your understanding
and enabling you to reconstruct things you can’t remember. But schematic
knowledge can sometimes hurt you, by promoting errors in perception and
memory. Moreover, the types of errors produced by schemata are quite predictable. As an example, imagine that you visit a dentist’s office, and this
one happens not to have any magazines in the waiting room. It’s likely that
you’ll forget this detail after a while, so what will happen when you later
try to recall your trip to the dentist? Odds are good that you’ll rely on schematic knowledge and “remember” that there were magazines (since, after all,
there usually are some scattered around a waiting room). In this way, your
286 •
C H A P T E R E I G H T Remembering Complex Events
recollection will make this dentist’s office seem more typical, more ordinary,
than it actually was.
Here’s the same point in more general terms. We’ve already said that schemata tell you what’s typical in a setting. Therefore, if you rely on schematic
knowledge to fill gaps in your recollection, you’ll fill those gaps with what’s
normally in place in that sort of situation. As a result, any reliance on schemata will make the world seem more “normal” than it really is and will make
the past seem more “regular” than it actually was.
This tendency toward “regularizing” the past has been documented in
many settings. The classic demonstration, however, comes from studies published long ago by British psychologist Frederick Bartlett. Bartlett presented
his participants with a story taken from the folklore of Native Americans
(Bartlett, 1932). When tested later, the participants did reasonably well in
recalling the gist of the story, but they made many errors in recalling the
particulars. The pattern of errors, though, was quite systematic: The details
omitted tended to be ones that made little sense to Bartlett’s British participants. Likewise, aspects of the story that were unfamiliar were often changed
into aspects that were more familiar; steps of the story that seemed inexplicable were supplemented to make the story seem more logical.
Overall, then, the participants’ memories seem to have “cleaned up” the
story they had read — making it more coherent (from their perspective), more
sensible. This is exactly what we would expect if the memory errors derived
from the participants’ attempts to understand the story and, with that,
their efforts toward fitting the story into a schematic frame. Elements that
fit within the frame remained in their memories (or could be reconstructed
later). Elements that didn’t fit dropped out of memory or were changed.
In the same spirit, consider the Brewer and Treyens study mentioned at
the start of this chapter — the study in which participants remembered seeing
shelves full of books, even though there were none. This error was produced
by schematic knowledge. During the event itself (while the participants were
sitting in the office), schematic knowledge told the participants that academic
offices usually contain many books, and this knowledge biased what the participants paid attention to. (If you’re already certain that the shelves contain
books, why should you spend time looking at the shelves? This would only
confirm something you already know — see Vo & Henderson, 2009.) Then,
when the time came to recall the office, participants used their schema to reconstruct what the office must have contained — a desk, a chair, and of course
lots of books. In this way, the memory for the actual office was eclipsed by
generic knowledge about what a “normal” academic office contains.
Likewise, think back to the misremembered plane crash and the related
studies of people remembering videos of other prominent events, even though
there were no videos of these events. Here, too, the memory errors distort
reality by making the past seem more regular, more typical, than it really
was. After all, people often hear about major news events via a television
broadcast or Internet coverage, and these reports usually include vivid video
footage. So here, too, the past as remembered seems to have been assimilated
TEST YOURSELF
3. W
hat is the evidence
that your understanding of an episode can
produce intrusion
errors?
4. What is the DRM
procedure, and what
results does this
procedure produce?
5. What is schematic
knowledge, and what
evidence tells us that
schematic knowledge
can help us — and
also can undermine
memory accuracy?
Memory Errors: A Hypothesis
•
287
into the pattern of the ordinary. The event as it unfolded was unusual, but the
event as remembered becomes typical of its kind — just as we would expect if
understanding and remembering were guided by our knowledge of the way
things generally unfold.
The Cost of Memory Errors
There’s clearly a “good news, bad news” quality to our discussion so far. On
the positive side, memory connections serve as retrieval paths, allowing you
to locate information in storage. The connections also enrich your understanding, because they tie each of your memories into a context provided
by other things you know. In addition, links to schematic knowledge enable
you to supplement your perception and recollection with well-informed (and
usually accurate) inference.
On the negative side, though, the same connections can undermine memory accuracy, and memory errors are troubling. As we’ve discussed in other
contexts, you rely on memory in many aspects of life, and it’s unsettling that
the memories you rely on may be wrong — misrepresenting how the past
unfolded.
Eyewitness Errors
EXONERATION OF
THE INNOCENT
Guy Miles spent more than
18 years in prison for an armed
robbery he did not commit.
He is one of the hundreds of
people who were convicted
in U.S. courts but then proven
innocent by DNA evidence.
Mistaken eyewitness evidence
accounts for more of these
false convictions than all other
causes combined. Note in
addition that these false
convictions typically involve
a “double error” — with someone innocent doing time in
jail, and the guilty person
walking around free.
288 •
In fact, we can easily find circumstances in which memory errors are large
in scale (not just concerned with minor details in the episode) and deeply
consequential. For example, errors in eyewitness testimony (e.g., identifying the wrong person as the culprit or misreporting how an event unfolded)
can potentially send an innocent person to jail and allow a guilty person
to go free.
How often do eyewitnesses make mistakes? One answer comes from U.S.
court cases in which DNA evidence, not available at the time of the trial,
shows that the courts had convicted people who were, in truth, not guilty.
There are now more than 350 of these exonerations, and the exonerees had
(on average) spent more than a dozen years in jail for crimes they didn’t
commit. Many of them were on death row, awaiting execution.
When closely examined, these cases yield a clear message. Some of
these men and women were convicted because of dishonest informants;
some because analyses of forensic evidence had been botched. But by far
the most common concern is eyewitness errors. In fact, according to most
analyses, eyewitness errors account for at least three quarters of these false
convictions — more than all other causes combined (e.g., Garrett, 2011;
Reisberg, 2014).
Cases like these make it plain that memory errors, including misidentifications, are profoundly important. We’re therefore led to ask: Are there ways
to avoid these errors? Or are there ways to detect the errors, so that we can
decide which memories are correct and which ones are not?
C H A P T E R E I G H T Remembering Complex Events
Planting False Memories
An enormous number of studies have examined eyewitness memory — the
sort of memory that police rely on when investigating crimes. In one of the
earliest procedures, Loftus and Palmer (1974) showed participants a series
of pictures depicting an automobile collision. Later, participants were asked
questions about the collision, but the questions were phrased in different ways for different groups. Some participants were asked, for example,
“How fast were the cars going when they hit each other?” A different group
was asked, “How fast were the cars going when they smashed into each
other?” The differences among these questions were slight, but had a substantial influence: Participants in the “hit” group estimated the speed to have
been 34 miles per hour; those in the “smashed” group estimated 41 miles
per hour — 20% higher (see Figure 8.3).
But what is critical comes next: One week later, the participants were
asked in a perfectly neutral way whether they had seen any broken glass
in the pictures. Participants who had initially been asked the “hit” question
tended to remember (correctly) that no glass was visible; participants who
had been asked the “smashed” question, though, often made this error. It
FIGURE 8.3
THE IMPACT OF LEADING QUESTIONS
Estimated speed (mph)
45
40
35
30
25
20
Contacted
Hit
Bumped
Collided
Smashed
Verb used
Witnesses who were asked how fast cars were going when they “hit” each
other reported (on average) a speed of 34 miles per hour. Other witnesses,
asked how fast the cars were going when they “smashed” into each other,
gave estimates 20% higher. When all participants were later asked whether
they’d seen broken glass at the scene, participants who’d been asked the
“smashed” question were more likely to say yes — even though there was no
broken glass.
(after loftus & palmer, 1974)
The Cost of Memory Errors
•
289
seems, therefore, that the change of just one word within the initial question
can have a significant effect — in this case, more than doubling the likelihood
of memory error.
In other studies, participants have been asked questions that contain overt
misinformation about an event. For example, they might be asked, “How
fast was the car going when it raced by the barn?” when, in truth, no barn
was in view. In still other studies, participants are exposed to descriptions of
the target event allegedly written by “other witnesses.” They might be told,
for example, “Here’s how someone else recalled the crime; does this match
what you recall?” Of course, the “other witness” descriptions contained some
misinformation, enabling researchers to determine if participants “pick up”
the false leads (e.g., Paterson & Kemp, 2006; also Edelson, Sharon, Dolan,
& Dudai, 2011). In other studies, researchers ask questions that require the
participants themselves to make up some bit of misinformation. For example,
participants could be asked, “In the video, was the man bleeding from his
knee or from his elbow after the fall?” Even though it was clear in the video
that the man wasn’t bleeding at all, participants are forced to choose one
of the two options (e.g., Chrobak & Zaragoza, 2008; Zaragoza, Payment,
Ackil, Drivdahl, & Beck, 2001).
These procedures differ in important ways, but they are all variations
on the same theme. In each case, the participant experiences an event and
then is exposed to a misleading suggestion about how the event unfolded.
Then some time is allowed to pass. At the end of this interval, the participant’s memory is tested. And in each of these variations, the outcome is the
same: A substantial number of participants — in some studies, more than
one third — end up incorporating the false suggestion into their memory of
the original event.
Of course, some attempts at manipulating memory are more successful, some less so. It’s easier, for example, to plant plausible memories rather
than implausible ones. (However, memories for implausible events can also
be planted — see Hyman, 2000; Mazzoni, Loftus, & Kirsch, 2001; Pezdek,
Blandon-Gitlin, & Gabbay, 2006; Scoboria, Mazzoni, Kirsch, & Jimenez,
2006; Thomas & Loftus, 2002.) Errors are also more likely if the post-event
information supplements what the person remembers, in comparison to contradicting what the person would otherwise remember. It’s apparently easier,
therefore, to “add to” a memory than it is to “replace” a memory (Chrobak &
Zaragoza, 2013). False memories are also more easily planted if the research
participants don’t just hear about the false event but, instead, are urged to
imagine how the suggested event unfolded. In one study, participants were
given a list of possible childhood events (going to the emergency room late at
night; winning a stuffed animal at a carnival; getting in trouble for calling 911)
and were asked to “picture each event as clearly and completely” as they
could. This simple exercise was enough to increase participants’ confidence
that the event had really occurred (Garry, Manning, Loftus, & Serman, 1996;
also Mazzoni & Memon, 2003; Sharman & Barnier, 2008; Shidlovski, Schul,
& Mayo, 2014).
290 •
C H A P T E R E I G H T Remembering Complex Events
Even acknowledging these variations, though, let’s emphasize the consistency of the findings. We can use subtle procedures (with slightly leading
questions) to plant false information in someone’s memory, or we can use
a more blatant procedure (demanding that the person make up the bogus
facts). We can use pictures, movies, or live events as the to-be-remembered
materials. In all cases, it’s remarkably easy to alter someone’s memory, with
the result that the past as the person remembers it can differ markedly from
the past as it really was. This is a widespread pattern, with numerous implications for how we think about the past and how we think about our reliance
on our own memories. (For more on research in this domain, see Carpenter
& Schacter, 2017; Cochran, Greenspan, Bogart, & Loftus, 2016; Frenda,
Nichols, & Loftus, 2011; Laney, 2012; Loftus, 2017; Rich & Zaragoza,
2016. For research documenting similar memory errors in children, see, e.g.,
Bruck & Ceci, 1999, 2009; Reisberg, 2014.)
Are There Limits on the Misinformation Effect?
The studies just described reflect the misinformation effect — a term referring
to memory errors that result from misinformation received after an event was
experienced. What sorts of memory errors can be planted in this way?
We’ve mentioned studies in which participants remember broken glass
when really there was none or remember a barn when there was no barn in
view. Similar procedures have altered how people are remembered — and so,
with just a few “suggestions” from the experimenter, participants remember
clean-shaven men as bearded, young people as old, and fat people as thin
(e.g., Christiaansen, Sweeney, & Ochalek, 1983; Frenda et al., 2011).
It’s remarkably easy to produce these errors — with just one word (“hit”
vs. “smashed”) being enough to alter an individual’s recollection. What happens, though, if we ramp up our efforts to plant false memories? Can we
create larger-scale errors? In one study, college students were told that the
investigators were trying to learn how different people remember the same
experience. The students were then given a list of events that (they were told)
had been reported by their parents; the students were asked to recall these
events as well as they could, so that the investigators could compare the students’ recall with their parents’ (Hyman, Husband, & Billings, 1995).
Some of the events on the list actually had been reported by the participants’ parents. Other events were bogus — made up by the experimenters.
One of the bogus events was an overnight hospitalization for a high fever; in
a different experiment, the bogus event was attending a wedding reception
and accidentally spilling a bowlful of punch on the bride’s family.
The college students were easily able to remember the genuine events (i.e.,
the events actually reported by their parents). In an initial interview, more than
80% of these events were recalled, but none of the students recalled the bogus
events. However, repeated attempts at recall changed this pattern. By a third
interview, 25% of the participants were able to remember the embarrassment
of spilling the punch, and many were able to supply the details of this (entirely
The Cost of Memory Errors
•
291
FIGURE 8.4
A
THE BALLOON RIDE THAT NEVER WAS
B
In this study, participants were shown a faked photo (Panel B) created from a real childhood snapshot
(Panel A). With this prompt, many participants were led to a vivid, detailed recollection of the balloon
ride — even though it never occurred!
fictitious) episode. Other studies have shown similar results. Participants have
been led to recall details of particular birthday parties that, in truth, they never
had (Hyman et al., 1995); or an incident of being lost in a shopping mall
even though this event never took place; or a (fictitious) event in which they
were the victim of a vicious animal attack (Loftus, 2003, 2004; also see, e.g.,
Chrobak & Zaragoza, 2008; Geraerts et al., 2009; Laney & Loftus, 2010).
Errors Encouraged through “Evidence”
Other researchers have taken a further step and provided participants with
“evidence” in support of the bogus memory. In one procedure, researchers
obtained a real childhood snapshot of the participant (see Figure 8.4A for an
example) and, with a few clicks of a computer mouse, created a fictitious picture like the one shown in Figure 8.4B. With this prompt, many participants
were led to a vivid, detailed recollection of the hot-air balloon ride — even
though it never occurred (Wade, Garry, Read, & Lindsay, 2002). Another
study used an unaltered photo showing the participants’ second-grade
class (see Figure 8.5 for an example). This was apparently enough to persuade participants that the experimenters really did have information about
their childhood. Therefore, when the experimenters “reminded” the participants about an episode of their childhood misbehavior, the participants
took this reminder seriously. The result: Almost 80% were able to “recall”
292 •
C H A P T E R E I G H T Remembering Complex Events
FIGURE 8.5
PHOTOGRAPHS CAN ENCOURAGE MEMORY ERRORS
In one study, participants were “reminded” of a (fictitious) stunt they’d pulled while in the second grade. Participants were much more likely to “remember” the stunt (and so more likely to develop a false memory) if the
experimenter showed them a copy of their actual second-grade class photo. Apparently, the photo convinced
the participants that the experimenter really did know what had happened, and this made the experimenter’s
(false) suggestion much more persuasive.
(lindsay et al., 2004)
the episode, often in detail, even though it had never happened (Lindsay,
Hagen, Read, Wade, & Garry, 2004).
False Memories, False Confessions
It is clear that people can sometimes remember entire events that never took
place. They sometimes remember emotional episodes (like being lost in a
shopping mall) that never happened. They can remember their own transgressions (spilling the punch bowl, misbehaving in the second grade), even
though these misdeeds never occurred.
One study pushed things still further, using a broad mix of techniques to
encourage false memories (Shaw & Porter, 2015). The interviewer repeatedly asked participants to recall an event that (supposedly) she had learned
about from their parents. She assured participants that she had detailed
information about the (fictitious) event, and she applied social pressure with
comments like “Most people are able to retrieve lost memories if they try
The Cost of Memory Errors
•
293
REMEMBERING
VISITORS
A substantial number of people
have vivid, elaborate memories
for an episode in which they
were abducted by space aliens.
They report the aliens’ medical examination of their human
captive; in some cases, they
describe being impregnated
by the aliens. Some people
regard these reports as proof
that our planet has been visited
by extraterrestrials. Most scientists, however, regard these
reports as false — as “memories” for an event that never
happened. On this interpretation, the abduction reports illustrate how wrong our memories
can sometimes be.
TEST YOURSELF
6. What is the misinformation effect?
Describe three
different procedures
that can produce
this effect.
7. Some people insist
that our memories
are consistently
accurate in remembering the gist, or overall
content, of an event;
when we make memory errors, they claim,
we make mistakes
only about the details
within an event. What
evidence allows us to
reject this claim?
294 •
hard enough.” She offered smiles and encouraging nods whenever participants showed signs of remembering the (bogus) target events. If participants
couldn’t recall the target events, she showed signs of disappointment and said
things like “That’s ok. Many people can’t recall certain events at first because
they haven’t thought about them for such a long time.” She also encouraged
participants to use a memory retrieval technique (guided imagery) that is
known to foster false memories.
With these (and other) factors in play, Shaw and Porter persuaded many
of their participants that just a few years earlier the participants had committed a crime that led to police contact. In fact, many participants seemed
able to remember an episode in which they had assaulted another person
with a weapon and had then been detained by the police. This felony never
happened, but many participants “recalled” it anyhow. Their memories were
in some cases vivid and rich with detail, and on many measures indistinguishable from memories known to be accurate.
Let’s be clear, though, that this study used many forms of influence and
encouragement. It takes a lot to pull memory this far off track! There has
also been debate over just how many of the participants in this study truly
developed false memories. Even so, the results show that it’s possible for a
large number of people to have memories that are emotionally powerful,
deeply consequential, and utterly false. (For discussion of Shaw and Porter’s
study, see Wade, Garry, & Pezdek, 2017. Also see Brewin & Andrews, 2017,
and then in response, Becker-Blease & Freyd, 2017; Lindsay & Hyman,
2017; McNally, 2017; Nash, Wade, Garry, Loftus, & Ost, 2017; Otgaar,
Merckelbach, Jelicic, & Smeets, 2017; and Scoboria & Mazzoni, 2017.)
C H A P T E R E I G H T Remembering Complex Events
COGNITION
outside the lab
“It’s Common Sense”
Psychology students sometimes get teased by
the risk of someday being kidnapped by space
their peers: “Why are you taking Psych courses?
aliens. In contrast, though, studies make it plain
It’s all a matter of common sense!” The same senti-
that just a word or two of leading can produce
ment can arise when psychologists testify in court
memory errors in roughly one third of the people
cases, with the goal of helping judges and juries
questioned. Surely, the danger of extraterrestrial
understand how memory functions — and how
abduction is much lower than this.
someone’s memory can be mistaken. Some judges,
Other commonsense beliefs are flatly wrong. For
however, refuse this testimony. In support of this
example, some people have the view that certain
refusal, they note that expert testimony is allowed
types of events are essentially immune to forget-
only if it will be helpful in deciding the case, and
ting. They speak about those events as somehow
the testimony won’t be helpful if it simply covers
“burned into the brain” and say things like “I’ll never
points that judge and jury already know. In legal
forget the events of 9/11” or “. . . the day I got mar-
jargon, the testimony is permitted only if it covers
ried” or “. . . what he looked like when he pulled the
topics “beyond the ken of the average juror.”
trigger.” However, the “burned into the brain” idea is
How should we think about these notions? Are
psychology’s claims about memory simply a con-
wrong, and investigators can often document largescale errors in these singular, significant memories.
firmation of common sense? Each of us, of course,
Additional examples are easy to find. These in-
has had a lifetime of experience working with and
clude the widely held view that someone’s degree of
relying on our memories; that experience has surely
certainty is a good index of whether his or her mem-
taught us a lot about how memory functions. Even
ory is accurate (this is true only in a narrow set of cir-
so, it’s easy to find widespread beliefs about mem-
cumstances); the common belief that our memories
ory that are incorrect. Often, these beliefs start
function just as a video recorder functions (not at all
with a kernel of truth but understate the actual
true); or the belief that hypnosis can allow someone
facts. For example, everyone knows that memo-
to recover long-lost memories (utterly false).
ries are sometimes inaccurate; people talk about
In fact, let’s note an irony here. Commonsense
their memories “playing tricks” on them. However,
beliefs about memory (or about psychology in
most people are astonished to learn how common
general) are sometimes sensible and sometimes
memory errors are and how large the errors can
not. If scientific research corrects a mistaken com-
sometimes be. Therefore, in relying on common
monsense belief, then obviously we’ve learned
sense, people (including judges and juries) prob-
something. But if the research turns out to con-
ably trust memory more than they should.
firm common sense, then here too we’ve learned
For example, in one study, college students
something — because we’ve learned that this is
were surveyed about their perceptions of vari-
one of the times when common sense is on track.
ous risks (Wilson & Brekke, 1994). These students
On that basis, we shouldn’t scoff at results that
were largely unconcerned about the risk of some-
“merely” confirm common sense, because these
one biasing their memory with leading questions;
results can be just as informative as results that
they regarded this risk as roughly equivalent to
truly surprise us.
The Cost of Memory Errors
•
295
Avoiding Memory Errors
Evidence is clear that people do make mistakes — at times, large mistakes — in
remembering the past. But people usually don’t make mistakes. In other
words, you generally can trust your memory, because more often than not
your recollection is detailed, long-lasting, and correct.
This mixed pattern, though, demands a question: Is there some way to
figure out when you’ve made a memory mistake and when you haven’t? Is there
a way to decide which memories you can rely on and which ones you can’t?
Memory Confidence
In evaluating memories, people rely heavily on expressions of certainty or
confidence. Specifically, people tend to trust memories that are expressed
with confidence. (“I distinctly remember her yellow jacket; I’m sure of it.”)
They’re more cautious about memories that are hesitant. (“I think she was
wearing yellow, but I’m not certain.”) We can see these patterns when people are evaluating their own memories (e.g., when deciding whether to take
action or not, based on a bit of recollection); we see the same patterns when
people are evaluating memories they hear from someone else (e.g., when
juries are deciding whether they can rely on an eyewitness’s testimony).
Evidence suggests, though, that a person’s degree of certainty is an uneven
indicator of whether a memory is trustworthy. On the positive side, there
are circumstances in which certainty and memory accuracy are highly correlated (e.g., Wixted, Mickes, Clark, Gronlund, & Roediger, 2015; Wixted
& Wells, 2017). On the negative side, though, we can easily find exceptions
to this pattern — including memories that are expressed with total certainty
(“I’ll never forget that day; I remember it as though it were yesterday”) but
that turn out to be entirely mistaken. In fact, we can find circumstances in
which there’s no correspondence at all between how certain someone says
she is, in recalling the past, and how accurate that recollection is likely to be.
As a result, if we try to categorize memories as correct or incorrect based on
someone’s confidence, we’ll often get it wrong. (For some of the evidence,
see Busey, Tunnicliff, Loftus, & Loftus, 2000; Hirst et al., 2009; Neisser &
Harsch, 1992; Reisberg, 2014; Wells & Quinlivan, 2009.)
How can this be? One reason is that a person’s confidence in a memory is
often influenced by factors that have no impact on memory accuracy. When
these factors are present, confidence can shift (sometimes upward, sometimes
downward) with no change in the accuracy level, with the result that any connection between confidence and accuracy can be strained or even shattered.
Participants in one study witnessed a (simulated) crime and later were asked
if they could identify the culprit from a group of pictures. Some of the participants were then given feedback — “Good, you identified the suspect”; others
weren’t. The feedback couldn’t possibly influence the accuracy of the identification, because the feedback arrived only after the identification had occurred.
But the feedback did have a large impact on how confident participants said
296 •
C H A P T E R E I G H T Remembering Complex Events
FIGURE 8.6
CONFIDENCE MALLEABILITY
In one study, participants first tried to identify a culprit
from a police lineup and then indicated (on a scale of
0 to 100) how confident they had been in their selection.
Some participants received no feedback about their choice;
others received feedback after making their selection
but before indicating their confidence level. The feedback
couldn’t possibly influence accuracy (because the selection had already been made), but it dramatically increased
confidence.
(after wells & bradfield, 1998)
Mean reported confidence
80
70
60
50
40
30
No feedback
they’d been when making their lineup selection (see Figure 8.6), and so, with
confidence inflated but accuracy unchanged, the linkage between confidence
and accuracy was essentially eliminated. (Wells & Bradfield, 1998; also see
Douglas, Neuschatz, Imrich, & Wilkinson, 2010; Semmler & Brewer, 2006;
Wells, Olson, & Charman, 2002, 2003; Wright & Skagerberg, 2007.)
Similarly, think about what happens if someone is asked to report on an
event over and over. The repetitions don’t change the memory content — and
so the accuracy of the report won’t change much from one repetition to
the next. However, with each repetition, the recall becomes easier and more
fluent, and this ease of recall seems to make people more confident that their
memory is correct. So here, too, accuracy is unchanged but confidence is
inflated — and thus there’s a gradual erosion, with each repetition, of the
correspondence between accuracy and confidence. (For more on the disconnection between accuracy and confidence, see, e.g., Bradfield Douglas &
Pavletic, 2012; Charman, Wells, & Joy, 2011.)
In many settings, therefore, we cannot count on confidence as a means
of separating accurate memories from inaccurate ones. In addition, other
findings tell us that memory errors can be just as emotional, just as vivid, as
accurate memories (e.g., McNally et al., 2004). In fact, research overall suggests that there simply are no indicators that can reliably guide us in deciding
which memories to trust and which ones not to trust. For now, it seems that
memory errors, when they occur, may often be undetectable.
“Good, you identified
our suspect”
Feedback condition
TEST YOURSELF
8. W
hat factors seem to
undermine the relationship between your
degree of certainty in a
memory and the likelihood that the memory
is accurate?
Forgetting
We’ve been discussing the errors people sometimes make in recalling the past,
but of course there’s another way your memory can let you down: Sometimes
you forget. You try to recall what was on the shopping list, or the name of
Forgetting
•
297
an acquaintance, or what happened last week, and you simply draw a blank.
Why does this happen? Are there things you can do to diminish forgetting?
The Causes of Forgetting
Let’s start with one of the more prominent examples of “forgetting” — which
turns out not to be forgetting at all. Imagine meeting someone at a party,
being told his name, and moments later realizing you don’t have a clue what
his name is — even though you just heard it. This common (and embarrassing)
experience is not the result of ultra-rapid forgetting. Instead, it stems from a
failure in acquisition. You were exposed to the name but barely paid attention to it and, as a result, never learned it in the first place.
What about “real” cases of forgetting — cases in which you once knew
the information but no longer do? For these cases, one of the best predictors
of forgetting (not surprisingly) is the passage of time. Psychologists use the
term retention interval to refer to the amount of time that elapses between
the initial learning and the subsequent retrieval; as this interval grows, you’re
likely to forget more and more of the earlier event (see Figure 8.7).
FIGURE 8.7
FORGETTING CURVE
100
90
Percentage retained
80
70
60
50
40
30
20
31 days
6 days
2 days
1 day
9 hours
60 mins
0 min
0
20 mins
10
Retention interval
The figure shows retention after various intervals since learning. The data
shown here are from classic work by Hermann Ebbinghaus, so the pattern is
often referred to as an “Ebbinghaus forgetting curve.” The actual speed of
forgetting (i.e., how “steep” the “drop-off” is) depends on how well learned
the material was at the start. Across most situations, though, the pattern is
the same — with the forgetting being rapid at first but then slowing down.
Mathematically, this pattern is best described by an equation framed in
terms of “exponential decay.”
298 •
C H A P T E R E I G H T Remembering Complex Events
One explanation for this pattern comes from the decay theory of forgetting,
which proposes rather directly that memories fade or erode with the passage
of time. Maybe this is because the relevant brain cells die off. Or maybe
the connections among memories need to be constantly refreshed — and if
they’re not refreshed, the connections gradually weaken.
A different possibility is that new learning somehow interferes with older
learning. This view is referred to as interference theory. According to this
view, the passage of time isn’t the direct cause of forgetting. Instead, the passage of time creates the opportunity for new learning, and it is the new learning that disrupts the older memories.
A third hypothesis blames retrieval failure. The idea here is that the “forgotten memory” is still in long-term storage, but the person trying to retrieve the
memory simply cannot locate it. This proposal rests on the notion that retrieval
from memory is far from guaranteed, and we argued in Chapter 7 that retrieval
is more likely if your perspective at the time of retrieval matches the perspective in place at the time of learning. If we now assume that your perspective is
likely to change as time goes by, we can make a prediction about forgetting:
The greater the retention interval, the greater the likelihood that your perspective has changed, and therefore the greater the likelihood of retrieval failure.
Which of these hypotheses is correct? It turns out that they all are. Memo­
ries do decay with the passage of time (e.g., Altmann & Schunn, 2012;
Wixted, 2004; also Hardt, Nader, & Nadel, 2013; Sadeh, Ozubko, Winocur,
& Moscovitch, 2016), so any theorizing about forgetting must include this
factor. But there’s also no question that a great deal of “forgetting” is retrieval
failure. This point is evident whenever you’re initially unable to remember
some bit of information, but then, a while later, you do recall that information. Because the information was eventually retrieved, we know that it
wasn’t “erased” from memory through either decay or interference. Your
initial failure to recall the information, then, must be counted as an example
of retrieval failure.
Sometimes retrieval failure is partial: You can recall some aspects of the
desired content, but not all. An example comes from the maddening circumstance in which you’re trying to think of a word but simply can’t come up
with it. The word is, people say, on the “tip of their tongue,” and following
this lead, psychologists refer to this as the TOT phenomenon. People experiencing this state can often recall the starting letter of the sought-after word
and approximately what it sounds like. So, for example, a person might
remember “it’s something like Sanskrit” in trying to remember “scrimshaw”
or “something like secant” in trying to remember “sextant” (Brown, 1991;
Brown & McNeill, 1966; Harley & Brown, 1998; James & Burke, 2000;
Schwartz & Metcalfe, 2011).
What about interference? In one early study, Baddeley and Hitch (1977)
asked rugby players to recall the names of the other teams they had played
against over the course of a season. The key here is that not all players
made it to all games (because of illness, injuries, or schedule conflicts).
This fact allows us to compare players for whom “two games back” means
Forgetting
•
299
two weeks ago, to players for whom “two games back” means four weeks
ago. In this way, we can look at the effects of retention interval (two
weeks vs. four) with the number of intervening games held constant. Likewise, we can compare players for whom the game a month ago was “three
games back” to players for whom a month ago means “one game back.”
Now, we have the retention interval held constant, and we can look at the
effects of intervening events. In this setting, Baddeley and Hitch reported
that the mere passage of time accounts for very little; what really matters
is the number of intervening events (see Figure 8.8). This is just what we
would expect if interference, and not decay, is the major contributor to
forgetting.
But why does memory interference occur? Why can’t the newly acquired
information coexist with older memories? The answer has several parts, but
one element is linked to issues we’ve already discussed: In many cases, newly
arriving information gets interwoven with older information, producing a
risk of confusion about which bits are old (i.e., the event you’re trying to
remember) and which bits are new (i.e., information that you picked up after
the event). In addition, in some cases, new information seems literally to
replace old information — much as you no longer save the rough draft of one
FIGURE 8.8
FORGETTING FROM INTERFERING EVENTS
100
Percentage of team names
recalled correctly
90
80
70
60
50
40
30
20
10
1
2
3
4
5 6 7 8 9 10 11 12 13 14 15 16 17 18
Number of intervening games played
Members of a rugby team were asked to recall the names of teams they had
played against. Overall, the broad pattern of the data shows that memory performance was powerfully influenced by the number of games that intervened
between the game to be recalled and the attempt to remember. This pattern fits
with an interference view of forgetting.
(after baddeley & hitch, 1977)
300 •
C H A P T E R E I G H T Remembering Complex Events
of your papers once the final draft is done. In this situation, the new information isn’t woven into the older memory; instead, it erases it.
Undoing Forgetting
Is there any way to undo forgetting and to recover seemingly lost memories?
One option, often discussed, is hypnosis. The idea is that under hypnosis a
person can “return” to an earlier event and remember virtually everything
about the event, including aspects the person didn’t even notice (much less
think about) at the time.
The reality, however, is otherwise. Hypnotized participants often do give
detailed reports of the target event, but not because they remember more; instead, they’re just willing to say more in order to comply with the hypnotist’s
instructions. As a result, their “memories” are a mix of recollection, guesses,
and inferences — and, of course, the hypnotized individual cannot tell which
of these are which (Lynn, Neuschatz, Fite, & Rhue, 2001; Mazzoni & Lynn,
2007; Spiegel, 1995).
On the positive side, though, there are procedures that do seem to diminish forgetting, including the so-called cognitive interview. This procedure was
A Drawings done by hypnotized adult told
that he was 6 years old
B Drawings done at age 6
HYPNOTIC AGE REGRESSION
In one study, participants were asked to draw a picture while mentally “regressed” to age 6. At first glance, their drawings
(see Panel A for an example) looked remarkably childlike. But when compared to the participants’ own drawings made
at that age (see Panel B for an example), it’s clear that the hypnotized adults’ drawings were much more sophisticated.
They represent an adult’s conception of what a childish drawing is, rather than being the real thing.
Forgetting
•
301
TEST YOURSELF
9. Explain the mechanisms hypothesized
by each of the three
major theories of
forgetting: decay,
interference, and
retrieval failure.
10. W
hat techniques or
procedures seem
ineffective as a means
of “un-doing” forgetting? What techniques
or procedures seem
to diminish or avoid
forgetting?
designed to help police in their investigations and, specifically, is aimed at
maximizing the quantity and accuracy of information obtained from eyewitnesses to crimes (Fisher & Schreiber, 2007; Memon, Meissner, & Fraser,
2010). The cognitive interview has several elements, including an effort
toward context reinstatement — steps that put witnesses back into the mindset they were in at the time of the crime. (For more on context reinstatement,
see Chapter 7.) In addition, the cognitive interview builds on the simple fact
that retrieval of memories from long-term storage is more likely if a suitable
cue is provided. The interview therefore offers a diverse set of retrieval cues
with the idea that the more cues provided, the greater the chance of finding
one that triggers the target memory.
The cognitive interview is quite successful, both in the laboratory and
in real crime investigations, producing more complete recollection without
compromising accuracy. This success adds to the argument that much of
what we call “forgetting” can be attributed to retrieval failure, and can be
undone simply by providing more support for retrieval.
Also, rather than undoing forgetting, perhaps we can avoid forgetting.
The key here is simply to “revisit” a memory periodically. Each “visit” seems
to refresh the memory, with the result that forgetting is much less likely. Researchers have examined this effect in several contexts, including one that’s
pragmatically quite important: Students often have to take exams, and confronting the material on an exam is, of course, an occasion in which students
“revisit” what they’ve learned. These revisits, we’ve just suggested, should
slow forgetting, and on this basis, taking an exam can actually help students
to hang on to the material they’ve learned. Several studies have confirmed
this “testing effect”: Students have better long-term retention for materials they were tested on, compared to materials they weren’t tested on. (See,
e.g., Carpenter, Pashler, & Cepeda, 2009; Halamish & Bjork, 2011; Healy,
Jones, Lalchandani, & Tack, 2017; Karpicke, 2012; Karpicke & Blunt, 2011;
McDaniel, Anderson, Derbish, & Morrisette, 2007; Pashler, Rohrer, Cepeda,
& Carpenter, 2007; Rowland, 2014.)
We might mention that similar effects can be observed if students test
themselves periodically, taking little quizzes that they’ve created on their
own. Related effects emerge if students are occasionally asked questions that
require a brief revisit to materials they’ve encountered (Brown, Roediger, &
McDaniel, 2014). In fact, that’s the reason why this textbook includes Test
Yourself questions; those questions will actually help readers to remember
what they’ve read!
Memory: An Overall Assessment
We’ve now seen that people sometimes recall with confidence events that
never took place, and sometimes forget information they’d hoped to remember. But we’ve also mentioned the positive side of things: how much people
can recall, and the key fact that your memory is accurate far more often than
not. Most of the time, it seems, you do recall the past as it truly was.
302 •
C H A P T E R E I G H T Remembering Complex Events
Perhaps most important, we’ve also suggested that memory’s “failings”
may simply be the price you pay in order to gain crucial advantages. For
example, we’ve argued that memory errors arise because the various episodes in your memory are densely interconnected with one another; it’s these
interconnections that allow elements to be transplanted from one remembered episode to another. But we’ve also noted that these connections have
a purpose: They’re the retrieval paths that make memory search possible.
Therefore, to avoid the errors, you would need to restrict the connections;
but if you did that, you would lose the ability to locate your own memories
within long-term storage.
The memory connections that lead to error also help you in other ways.
Our environment, after all, is in many ways predictable, and it’s enormously useful for you to exploit that predictability. There’s little point,
for example, in scrutinizing a kitchen to make sure there’s a stove in the
room, because in the vast majority of cases there is. So why take the time
to confirm the obvious? Likewise, there’s little point in taking special note
that, yes, this restaurant does have menus and, yes, people in the restaurant are eating and not having their cars repaired. These, too, are obvious
points, and it would be a waste of effort to give them special notice.
On these grounds, reliance on schematic knowledge is a good thing. Schemata guide your attention to what’s informative in a situation, rather than
what’s self-evident (e.g., Gordon, 2006), and they guide your inferences at
the time of recall. If this use of schemata sometimes leads you astray, that’s a
small price to pay for the gain in efficiency that schemata allow. (For similar
points, see Chapter 4.)
In the same way, the blurring together of episodes may be a blessing, not
a problem. Think, for example, about all the times when you’ve been with
a particular friend. These episodes are related to one another in an obvious
way, and so they’re likely to become interconnected in your memory. This
will cause difficulties if you want to remember which episode is which and
whether you had a particular conversation in this episode or in that one.
But rather than lamenting this, maybe we should celebrate what’s going on
here. Because of the “interference,” all the episodes will merge together in
your memory, so that what resides in memory is one integrated package,
containing all of your knowledge about your friend. As a result, rather than
complaining about memory confusion, we should rejoice over the memory
integration and “cross-referencing.”
In all of these ways, then, our overall assessment of memory can be rather
upbeat. We have, to be sure, discussed a range of memory errors, but these
errors are in most cases a side product of mechanisms that otherwise help
you — to locate your memories within storage, to be efficient in your contact
with the world, and to form general knowledge. Thus, even with the errors,
even with forgetting, it seems that human memory functions in a way that
serves us extraordinarily well. (For more on the benefits produced by memory’s apparent limitations, see Howe, 2011; Nørby, 2015; Schacter, Guerin,
& St. Jacques, 2011.)
TEST YOURSELF
11.Explain why the
mechanisms that
produce memory
errors may actually be
mechanisms that help
us in important ways.
Memory: An Overall Assessment
•
303
Autobiographical Memory
Most of the evidence in Chapters 6 and 7 was concerned with memory for
simple stimuli — such as word lists or short sentences. In this chapter, we’ve
considered memories for more complex materials, and this has drawn our
attention to the ways in which your knowledge (whether knowledge of a
general sort or knowledge about related episodes) can both improve memory
and also interfere with it.
In making these points, we’ve considered memories in which the research
participant was actually involved in the remembered episode, and not just an
external witness (e.g., the false memory that he committed a felony). We’ve
also looked at studies that involved memories for emotional events (e.g., the
plane crash discussed at the chapter’s start) and memory over the very long
term (e.g., memories for childhood events “planted” in adult participants).
Do these three factors — involvement in the remembered event, emotion,
and long delay — affect how or how well someone remembers? These factors
are surely relevant to the sorts of remembering people do outside the laboratory, and all three are central for autobiographical memory. This is the memory that each of us has for the episodes and events of our lives, and this sort
of memory plays a central role in shaping how each of us thinks about ourselves and, therefore, how we behave. (For more on the importance of autobiographical memory, see Baddeley, Aggleton, & Conway, 2002; Prebble, Addis,
& Tippett, 2013; Steiner, Thomsen, & Pillemer, 2017. For more on the distinction between the types of memory, including biological differences between
autobiographical memory and “lab memory,” see Cabeza & St. Jacques, 2007;
Hodges & Graham, 2001; Kopelman & Kapur, 2001; Tulving, 1993, 2002.)
Let’s explore how the three factors we’ve mentioned, each seemingly central for autobiographical memory, influence what we remember.
Memory and the Self
Having some involvement in an event (as opposed to passively witnessing it)
turns out to have a large effect on memory, because, overall, information
relevant to the self is better remembered than information that’s not selfrelevant — a pattern known as the “self-reference effect” (e.g., Symons &
Johnson, 1997; Westmacott & Moscovitch, 2003). This effect emerges in many
forms, including an advantage in remembering adjectives that apply to you
relative to adjectives that don’t, better memory for names of places you have
visited relative to names of places you’ve never been, and so on (see Figure 8.9).
But here, too, we can find memory errors, in part because your “memory”
for your own life is (just like other memories) a mix of genuine recall and
some amount of schema-based reconstruction. For example, consider the
fact that most adults believe they’ve been reasonably consistent, reasonably
stable, over their lifetimes. They believe, in other words, that they’ve always
been pretty much the same as they are now. This idea of consistency is part of
their self-schema — the set of interwoven beliefs and memories that constitute
304 •
C H A P T E R E I G H T Remembering Complex Events
FIGURE 8.9
SELF-REFERENCING AND THE BRAIN
Words in relation to another person
Words in relation to its printed format
MPFC activated
during self-referential
condition.
Words in relation to self
0.2
Signal change (%)
0.1
0
–0.1
–0.2
–0.3
–0.4
–5
0
5
10
15
20
Time (s)
You are more likely to remember words that refer to you, in comparison to words in other categories. Here, participants were asked to judge adjectives in three conditions: answering questions like “Does this word describe
the president?” or “Is this word printed in capital letters?” or “Does this word describe you?” Data from fMRI
recordings showed a distinctive pattern of processing when the words were “self-referential.” Specifically, selfreferential processing is associated with activity in the medial prefrontal cortex (MPFC). This extra processing is
part of the reason why self-referential words are better remembered.
(after kelley et al., 2002)
people’s knowledge about themselves. When the time comes to remember
the past, therefore, people will rely to some extent on this belief in their own
consistency, so they’ll reconstruct their history in a biased way — one that
maximizes the (apparent) stability of their lives. As a result, people often misremember their past attitudes and past romantic relationships, unwittingly
distorting their personal history in a way that makes the past look more
like the present than it really was. (See Conway & Ross, 1984; Holmberg &
Homes, 1994. For related results, see Levine, 1997; Marcus, 1986; McFarland
& Buehler, 2012; Ochsner & Schacter, 2000; Ross & Wilson, 2003.)
It’s also true that most of us would like to have a positive view of ourselves, including a positive view of how we’ve acted in the past. This, too, can
shape memory. As one illustration, Bahrick, Hall, and Berger (1996) asked
college students to recall their high school grades as accurately as they could,
and the data showed a clear pattern of self-service. When students forgot a
good grade, their (self-serving) reconstruction led them to the (correct) belief
Autobiographical Memory
•
305
that the grade must have been a good one; consistent with this, 89% of the A’s
were correctly remembered. But when students forgot a poor grade, reconstruction led them to the (false) belief that the grade must have been okay; as
a result, only 29% of the D’s were correctly recalled. (For other mechanisms
through which motivation can color autobiographical recall, see Conway &
Holmes, 2004; Conway & Pleydell-Pearce, 2000; Molden & Higgins, 2012.)
Memory and Emotion
Another factor important for autobiographical memory is emotion. Many
of your life experiences are of course emotional, making you feel happy, or
sad, or angry, or afraid, and in general emotion helps you to remember. One
reason is emotion’s impact on memory consolidation — the process through
which memories are biologically “cemented in place.” (See Hardt, Einarsson,
& Nader, 2010; Wang & Morris, 2010; although also see Dewar, Cowan, &
Della Sala, 2010.)
Whenever you experience an event or gain new knowledge, your memory
for this new content is initially fragile and is likely represented in the brain
via a pattern of neural activation. Over the next few hours, though, various
biological processes stabilize this memory and put it into a more enduring
form. This process — consolidation — takes place “behind the scenes,” without you thinking about it, but it’s crucial. If the consolidation is interrupted
for some reason (e.g., because of extreme fatigue or injury), no memory is established and recall later will be impossible. (That’s because there’s no information in memory for you to retrieve; you can’t read text off a blank page!)
A number of factors can promote consolidation. For example, evidence is increasing that key steps of consolidation take place while you’re
asleep — and so a good night’s rest actually helps you, later on, to remember
things you learned while awake the day before. (See Ackermann & Rasch,
2014; Giuditta, 2014; Rasch & Born, 2013; Tononi & Cirelli, 2013; Zillmer,
Spiers, & Culbertson, 2008.)
Also, there’s no question that emotion enhances consolidation. Specifically, emotional events trigger a response in the amygdala, and the amygdala
in turn increases activity in the hippocampus. The hippocampus is, as we’ve
seen, crucial for getting memories established. (See Chapter 7; for reviews of
emotion’s biological effects on memory, see Buchanan, 2007; Hoschedidt,
Dongaonkar, Payne, & Nadel, 2010; Joels, Fernandez, & Roosendaal, 2011;
Kensinger, 2007; LaBar, 2007; LaBar & Cabeza, 2006; Yonelinas & Ritchey,
2015. For a complication, though, see Figure 8.10.)
Emotion also shapes memory through other mechanisms. An event that’s
emotional is likely to be important to you, virtually guaranteeing that you’ll pay
close attention as the event unfolds, and we know that attention and thoughtful processing help memory. Moreover, you tend to mull over emotional events
in the minutes (or hours) following the event, and this is tantamount to memory rehearsal. For all these reasons, it’s not surprising that emotional events
are well remembered (Reisberg & Heuer, 2004; Talmi, 2013).
306 •
C H A P T E R E I G H T Remembering Complex Events
FIGURE 8.10
I NDIVIDUAL DIFFERENCES IN
EPISODIC MEMORY
BB
EE
KB
BK
HG
NL
CC
JL
SC
Group
Researchers have made enormous progress in explaining the brain mechanisms that support memory. One complication, though, is that the brain
mechanisms may differ from one individual to the next. This figure shows
data from nine different people (and then an average of the nine) engaged
in a task requiring the retrieval of episodic memories. As you can see, the
pattern of brain activation differs somewhat from person to person.
( after miller et al ., 2002)
Let’s note, though, that emotion doesn’t just influence how well you remember; it also influences what you remember. Specifically, in many settings,
emotion seems to produce a “narrowing” of attention, so that all of your
attention will be focused on just a few aspects of the scene (Easterbrook,
1959). This narrowing helps guarantee that these attended aspects will be
firmly placed into memory, but it also implies that the rest of the event,
excluded from the narrowed focus, won’t be remembered later (e.g., Gable
& Harmon-Jones, 2008; Reisberg & Heuer, 2004; Steblay, 1992).
What exactly you’ll focus on, though, may depend on the specific emotion.
Different emotions lead you to set different goals: If you’re afraid, your goal
is to escape; if you’re angry, your goal is to deal with the person or issue that’s
Autobiographical Memory
•
307
made you angry; if you’re happy, your goal may be to relax and enjoy! In each
case, you’re more likely to pay attention to aspects of the scene directly rele­
vant to your goal, and this will color how you remember the emotional event.
(See Fredrickson, 2000; Harmon-Jones, Gable, & Price, 2013; Huntsinger,
2012, 2013; Kaplan, Van Damme, & Levine, 2012; Levine & Edelstein, 2009.)
Flashbulb Memories
One group of emotional memories seems special. These are the so-called
flashbulb memories — memories of extraordinary clarity, typically for highly
emotional events, retained despite the passage of many years. When Brown
and Kulik (1977) introduced the term “flashbulb memory,” they pointed to
the memories people had of the moment in 1963 when they first heard that
President Kennedy had been assassinated. In the Brown and Kulik study,
people interviewed more than a decade after that event remembered it
“as though it were yesterday,” and many participants were certain they’d
never forget that awful day. Moreover, participants’ recollection was quite
detailed — with people remembering where they were at the time, what they
were doing, and whom they were with. Indeed, many participants were able
to recall the clothing worn by people around them, the exact words uttered,
and the like.
Many other events have also produced flashbulb memories. For example,
most Americans can clearly recall where they were when they heard about
the attack on the World Trade Center in 2001; many people vividly remember what they were doing in 2009 when they heard that Michael Jackson had
died; many Italians have clear memories of their country’s victory in the 2006
World Cup; and so on. (See Pillemer, 1984; Rubin & Kozin, 1984; also see
Weaver, 1993; Winograd & Neisser, 1993.)
Remarkably, though, these vivid, high-confidence memories can contain
substantial errors. Thus, when people say, “I’ll never forget that day . . .” they’re
sometimes wrong. For example, Hirst et al. (2009) interviewed more than
3,000 people soon after the September 11 attack on the World Trade Center,
asking how they first heard about the attack; who brought them the news; and
what they were doing at the time. When these individuals were re-interviewed
a year later, however, more than a third (37%) provided a substantially different account. Even so, the participants were strongly confident in their recollection (rating their degree of certainty, on a 1-to-5 scale, at an average of 4.4).
The outcome was the same for participants interviewed three years after the
attack — with 43% offering different accounts from those they had given initially. (For similar data, see Neisser & Harsch, 1992; also Hirst & Phelps,
2016; Rubin & Talarico, 2007; Schmidt, 2012; Talarico & Rubin, 2003.)
Other data, though, tell a different story, suggesting that some flashbulb
memories are entirely accurate. Why should this be? Why are some flashbulb
events remembered well, while others aren’t? The answer involves several
factors, including how, how often, and with whom someone discusses the flashbulb event. In many cases, this discussion may encourage people to “polish”
their reports — so that they’re offering their audience a “better,” more interesting narrative. After a few occasions of telling and re-telling this version of the
308 •
C H A P T E R E I G H T Remembering Complex Events
FLASHBULB MEMORIES
People often have especially clear and long-lasting memories for events like first hearing about Princess Diana’s
death in 1997, the attack on the World Trade Center in September 2001, or the news of Michael Jackson’s death in
2009. These memories — called “flashbulb memories” — are vivid and compelling, but they are not always accurate.
event, the new version may replace the original memory. (For more on these
issues, see Conway et al., 1994; Hirst et al., 2009; Luminet & Curci, 2009;
Neisser, Winograd, & Weldon, 1991; Palmer, Schreiber, & Fox, 1991; Tinti,
Schmidt, Sotgiu, Testa, & Curci, 2009; Tinti, Schmidt, Testa, & Levine, 2014.)
Notice, then, that an understanding of flashbulb memories requires us to
pay attention to the social aspects of remembering. In many cases, people
“share” memories with one another (and so, for example, I tell you about
my vacation, and you tell me about yours). Likewise, in the aftermath of
an important event, people often compare their recollections. (“Did you see
how he ran when the alarm sounded!?”) In all cases, people are likely to alter
their accounts in various ways, to allow for a better conversation. They may,
for example, leave out mundane bits, or add bits to make their account more
interesting or to impress their listeners. These new points about how the event
is described will, in turn, often alter the way the event is later remembered.
In addition, people sometimes “pick up” new information in these conversations — if, for example, someone who was present for the same event noticed
a detail that you missed. Often, this new information will be absorbed into
other witnesses’ memory — a pattern sometimes referred to as “co-witness
contamination.” Let’s note, though, that sometimes another person who witnessed the event will make a mistake in recalling what happened, and, after
conversation, other witnesses may absorb this mistaken bit into their own
recollection (Hope, Gabbert, & Fraser, 2013). In this way, conversations after
Autobiographical Memory
•
309
an event can sometimes have a positive impact on the accuracy and content
of a person’s eventual report, and sometimes a negative impact.
For all these reasons, then, it seems that “remembering” is not an activity
shaped only by the person who holds the memory, and exploring this point will
be an important focus for future research. (For early discussion of this broad
issue, see Bartlett, 1932. For more recent discussion, see Choi, Kensinger, &
Rajaram, 2017; Gabbert & Hope, 2013; Roediger & Abel, 2015.)
Returning to flashbulb memories, though, let’s not lose track of the fact
that the accuracy of these memories is uneven. Some flashbulb memories
are marvelously accurate; others are filled with error. Therefore, the commonsense idea that these memories are somehow “burned into the brain,”
and thus always reliable, is surely mistaken. In addition, let’s emphasize that
from the point of view of the person who has a flashbulb memory, there’s no
detectable difference between an accurate flashbulb memory and an inaccurate
one: Either one will be recalled with great detail and enormous confidence. In
each case, the memory can be intensely emotional. Apparently, memory errors
can occur even in the midst of our strongest, most vivid recollections.
Traumatic Memories
Flashbulb memories usually concern events that were strongly emotional.
Sadly, though, we can also find cases in which people experience truly extreme
emotion, and this leads us to ask: How are traumatic events remembered? If
someone has witnessed wartime atrocities, can we count on the accuracy of
their testimony in a war-crimes trial? If someone suffers through the horrors
of a sexual assault, will the painful memory eventually fade?
Evidence suggests that most traumatic events are well remembered for
many years. In fact, victims of atrocities often seem plagued by a cruel
enhancement of memory, leaving them with extra-vivid and long-lived recollections of the terrible event (e.g., Alexander et al., 2005; Goodman et al.,
2003; Peace & Porter, 2004; Porter & Peace, 2007; Thomsen & Berntsen,
2009). As a result, people who have experienced trauma sometimes complain
about having “too much” memory and wish they remembered less.
This enhanced memory can be understood in terms of a mechanism we’ve
already discussed: consolidation. This process is promoted by the conditions that accompany bodily arousal, including the extreme arousal typically
present in a traumatic event (Buchanan & Adolphs, 2004; Hamann, 2001;
McGaugh, 2015). But this doesn’t mean that traumatic events are always
well remembered. There are, in fact, cases in which people who’ve suffered
through extreme events have little or no recall of their experience (e.g.,
Arrigo & Pezdek, 1997). We can also sometimes document substantial errors
in someone’s recall of a traumatic event (Paz-Alonso & Goodman, 2008).
What factors are producing this mixed pattern? In some cases, traumatic
events are accompanied by sleep deprivation, head injuries, or substance
abuse, each of which can disrupt memory (McNally, 2003). In other cases,
the memory-promoting effects of arousal are offset by the complex memory effects of stress. The key here is that the experience of stress sets off a
310 •
C H A P T E R E I G H T Remembering Complex Events
cascade of biological reactions. These reactions produce changes throughout
the body, and the changes are generally beneficial, helping the organism to
survive the stressful event. However, the stress-produced changes are disruptive to some biological functions, and this can lead to a variety of problems
(including medical problems caused by stress).
How does the mix of stress reactions influence memory? The answer is
complicated. Stress experienced at the time of an event seems to enhance
memory for materials directly relevant to the source of the stress, but has
the opposite effect — undermining memory — for other aspects of the event
(Shields, Sazma, McCullough, & Yonelinas, 2017). Also, stress experienced
during memory retrieval interferes with memory, especially if the target information was itself emotionally charged.
How does all this play out in situations away from the laboratory? One
line of evidence comes from a study of soldiers who were undergoing survival training. As part of their training, the soldiers were deprived of sleep
and food, and they went through a highly realistic simulation of a prisonerof-war interrogation. One day later, the soldiers were asked to identify the
interrogator from a lineup. Despite the extensive (40-minute) face-to-face
encounter with the interrogator and the relatively short (one-day) retention
interval, many soldiers picked the wrong person from the lineup. Soldiers
who had experienced a moderate-stress interrogation picked the wrong
person from a live lineup 38% of the time; soldiers who had experienced a
high-stress interrogation (one that included a physical confrontation) picked
the wrong person 56% of the time if tested with a live lineup, and 68% of the
time if tested with a photographic lineup. (See Morgan et al., 2004; also see
Deffenbacher, Bornstein, Penrod, & McCorty, 2004; Hope, Lewinski, Dixon,
Blocksidge, & Gabbert, 2012; Valentine & Messout, 2008.)
Repression and “Recovered” Memories
Some authors argue in addition that people defend themselves against extremely painful memories by pushing these memories out of awareness. Some
writers suggest that the painful memories are “repressed”; others use the term
“dissociation” to describe this self-protective mechanism. No matter what
terms we use, the idea is that these painful memories (including, in many
cases, memories for childhood abuse) won’t be consciously available but will
still exist in a person’s long-term storage and in suitable circumstances can be
“recovered” — that is, made conscious again. (See, for discussion, Belli, 2012;
Freyd, 1996, 1998; Terr, 1991, 1994.)
Most memory researchers, however, are skeptical about this proposal. As
one consideration, painful events — including events that seem likely candidates
for repression — seem typically to be well remembered, and this is the opposite
of what we would expect if a self-protective mechanism was in place. In addition, some of the abuse memories reported as “recovered” may, in fact, have
been remembered all along, and so they provide no evidence of repression or
dissociation. In these cases, the memories had appeared to be “lost” because the
person refused to discuss these memories for many years; “recovery” of these
Autobiographical Memory
•
311
memories simply reflects the fact that the person is at last willing to talk about
them. This sort of “recovery” can be extremely consequential — emotionally
and legally — but doesn’t tell us anything about how memory works.
Sometimes, though, memories do seem to be genuinely lost for a while
and then recovered. But this pattern may not reveal the operation (and, eventually, the “lifting”) of repression or dissociation. Instead, this pattern may
be the result of retrieval failure — a mechanism that can “hide” memories
for periods of time, only to have them reemerge once a suitable retrieval
cue is available. Here, too, the recovery may be of enormous importance
for the person who is finally remembering the long-lost episodes; but again,
this merely confirms the role of an already-documented memory mechanism,
with no need for theorizing about repression.
In addition, we need to acknowledge the possibility that at least some
recovered memories may, in fact, be false memories. After all, we know that
false memories occur and that they’re more likely when someone is recalling
the distant past than when one is trying to remember recent events. It’s also
relevant that many recovered memories emerge only with the assistance of a
therapist who is genuinely convinced that a client’s psychological problems
stem from long-forgotten episodes of childhood abuse. Even if therapists
scrupulously avoid leading questions, their expectations might still lead them
to shape their clients’ memory in other ways — for example, by giving signs
of interest or concern if the clients hit on the “right” line of exploration, by
spending more time on topics related to the alleged memories, and so on. In
these ways, the climate within a therapeutic session could guide the client
toward finding exactly the “memories” the therapist expects to find.
Overall, then, the idea of a self-protective mechanism “hiding” painful
memories from view is highly controversial. Some psychologists (often, those
working in a mental health specialty) insist that they routinely observe this
sort of self-protection, and other psychologists (generally, memory researchers) reject the idea that memories can be hidden in this way. It does seem
clear, however, that at least some of these now-voiced memories are accurate and provide evidence for terrible crimes. As in all cases, though, the
veracity of recollection cannot be taken for granted. This warning is important in evaluating any memory, but especially so for anyone wrestling with
traumatic recollection. (For discussions of this difficult — and sometimes
angrily debated — issue, see, among others, Belli, 2012; Brewin &Andrews,
2014, 2016; Dalenberg et al., 2012; Geraerts et al., 2009; Giesbrecht,
Lynn, Lilienfeld, & Merckelbach, 2008; Kihlstrom, 2006; Küpper, Benoid,
Dalgleish, & Anderson, 2014; Loftus, 2017; Ost, 2013; Patihis, Lilienfeld,
Ho, & Loftus, 2014; Pezdek & Blandon-Gitlin, 2017.)
Long, Long-Term Remembering
In the laboratory, a researcher might ask you to recall a word list you read just
minutes ago or a film you saw a week ago. Away from the lab, however, people routinely try to remember events from years — perhaps decades — back.
312 •
C H A P T E R E I G H T Remembering Complex Events
We’ve mentioned that these longer retention intervals are generally associated with a greater amount of forgetting. But, impressively, memories from
long ago can sometimes turn out to be entirely accurate.
In an early study, Bahrick, Bahrick, and Wittlinger (1975; also Bahrick,
1984; Bahrick & Hall, 1991) tracked down the graduates of a particular high
school — people who had graduated in the previous year, and the year before,
and the year before that, and ultimately, people who had graduated 50 years
earlier. These alumni were shown photographs from their own year’s high
school yearbook, and for each photo they were given a group of names and
had to choose the name of the person shown in the picture. The data for this
“name-matching” task show remarkably little forgetting; performance was
approximately 90% correct if tested 3 months after graduation, the same
after 7 years, and the same after 14 years. In some versions of the test, performance was still excellent after 34 years (see Figure 8.11).
FIGURE 8.11
MEMORY OVER THE VERY LONG TERM
Percentage of correct answers
100
80
60
40
Name matching
20
25 yr 10 mo
34 yr 1 mo
47 yr 7 mo
14 yr 6 mo
7 yr 5 mo
3 yr 10 mo
1 yr 11 mo
9.3 mo
0
3.3 mo
Picture cueing
Time since graduation
When people were tested for how well they remembered names and faces of
their high school classmates, their memory was remarkably long-lasting. In
the name-matching task, participants were given a group of names and had
to choose the right one. In the picture-cueing task, participants had to come
up with the names on their own. In both tasks, the data show a sharp dropoff after 47 years, but it is unclear whether this reflects an erosion of memory
or a more general drop-off in performance caused by the normal process
of aging.
(after bahrick, bahrick, & wittlinger, 1975)
Autobiographical Memory
•
313
As a different example, what about the material you’re learning right
now? Five years from now, will you still remember what you’ve learned?
How about a decade from now? Conway, Cohen, and Stanhope (1991, 1992)
explored these questions, testing students’ retention of a cognitive psychology
course taken years earlier. The results echo the pattern we’ve already seen.
Some forgetting of names and specific concepts was observed during the first
3 years after the course. After the third year, however, performance stabilized,
so that students tested after 10 years still remembered a fair amount — in
fact, just as much as students tested after 3 years (see Figure 8.12).
In an earlier section, we argued that the retention interval is crucial for
memory and that memory gets worse as times goes by. The data now in front
of us, though, indicate that how much the interval matters — that is, how
quickly memories “fade” — may depend on how well established the memories were in the first place. The high school students in the Bahrick et al. study
had seen their classmates day after day, for (perhaps) several years. They
therefore knew their classmates’ names very, very well — and this is why the
passage of time had only a slight impact on their memories for the names.
Likewise, students in the Conway et al. study had apparently learned their
psychology quite well — and so they retained what they’d learned for a very
long time. In fact, we first met this study in Chapter 6, when we mentioned
that students’ grades in the course were good predictors of how much the
students would still remember many years after the course was done. Here,
too, the better the original learning, the slower the forgetting.
314 •
80
70
Concepts
60
Names
50
Chance
C H A P T E R E I G H T Remembering Complex Events
Retention interval Times
10 yr 5 mo
9 yr 5 mo
8 yr 5 mo
7 yr 5 mo
6 yr 5 mo
5 yr 5 mo
4 yr 5 mo
3 yr 5 mo
3 yr 3 mo
2 yr 3 mo
1 yr 3 mo
40
3 mo
Participants in this study were
quizzed about material they had
learned in a college course taken
as recently as three months ago
or as far back as a decade ago.
The data showed some forgetting,
but then performance leveled
off; memory seemed remarkably
stable from three years onward.
Note that in a recognition task,
memory is probed with “familiaror-not” questions, so someone
with no memory, responding at
random, would get 50% right just
by chance.
(after conway, cohen, & stanhope, 1991)
90
Mean percentage correctly recognized
FIGURE 8.12 LONG-TERM
RETENTION OF COURSE
MATERIALS
We can maintain our claim, therefore, that the passage of time is the
enemy of memory: Longer retention intervals produce lower levels of recall.
However, if the material is very well learned at the start, and also if you
periodically “revisit” the material, you can dramatically diminish the impact
of the passing years.
How General Are the Principles
of Memory?
TEST YOURSELF
12. W
hat is memory
consolidation?
13. What is a flashbulb
memory? Are
flashbulb memories
distinctive in how
accurate they seem
to be?
There is certainly more to be said about autobiographical memory. For
example, it can’t be surprising that people tend to remember significant
turning points in their lives and often use these turning points as a means
of organizing their autobiographical recall (Enz & Talarico, 2015; Rubin
& Umanath, 2015). Perhaps related, there are also memory patterns associated with someone’s age. Specifically, most people recall very little from
the early years of childhood (before age 3 or so; e.g., Akers et al., 2014;
Bauer, 2007; Hayne, 2004; Howe, Courage, & Rooksby, 2009; Morrison &
Conway, 2010). In contrast, people generally have clear and detailed memories of their late adolescence and early adulthood, a pattern known as the
“reminiscence bump.” (See Figure 8.13; Conway & Haque, 1999; Conway,
Wang, Hanyu, & Haque 2005; Dickson, Pillemer, & Bruehl, 2011; Koppel &
Rubin, 2016; Rathbone, Moulin, & Conway, 2008; Rathbone, O’Connor, &
Moulin, 2017.) As a result, for many Americans, the last years of high school
35
Percentage of Memories
30
Japan
China
Bangladesh
US
UK
All
25
20
15
10
5
0
5
10
15
20
25
30
35
40
45
Age at Encoding (in 5-year bins)
50
55
60
FIGURE 8.13 THE
LIFESPAN RETRIEVAL
CURVE
Most people have few memories
of their early childhood (roughly
from birth to age 3 or 4); this
pattern is referred to as “childhood amnesia.” In contrast, the
period from age 10 to 30 is well
remembered, producing a pattern called the “reminiscence
bump.” This “bump” has been
observed in multiple studies and
in diverse cultures; events from
this time in young adulthood are
often remembered in more detail
(although perhaps less accurately)
than more recent events.
How General Are the Principles of Memory?
•
315
and the years they spend in college are likely to be the most memorable
periods of their lives.
But in terms of the broader themes of this chapter, where does our brief
survey of autobiographical memory leave us? In many ways, this form
of memory is similar to other sorts of remembering. Autobiographical
memories can last for years and years, but so can memories that don’t refer
directly to your own life. Autobiographical remembering is far more likely
if the person occasionally revisits the target memories; these rehearsals dramatically reduce forgetting. But the same is true in non-autobiographical
remembering.
Autobiographical memory is also open to error, just as other forms of
remembering are. We saw this in cases of flashbulb memories that turn out
to be false. We’ve also seen that misinformation and leading questions can
plant false autobiographical memories — about birthday parties that never
happened and trips to the hospital that never took place (also see Brown
& Marsh, 2008). Misinformation can even reshape memories for traumatic
events, just as it can alter memories for trivial episodes in the laboratory
(Morgan, Southwick, Steffan, Hazlett, & Loftus, 2013; Paz-Alonso &
Goodman, 2008).
These facts strengthen a claim that has been emerging in our discussion
over the last three chapters: Certain principles seem to apply to memory in
general, no matter what is being remembered. All memories depend on connections. The connections promote retrieval. The connections also facilitate
interference, because they allow one memory to blur into another. The connections can fade with the passage of time, producing memory gaps, and
the gaps are likely to be filled via reconstruction based on generic knowledge. All these things seem to be true whether we’re talking about relatively
recent memories or memories from long ago, emotional memories or memories of calm events, memories for complex episodes or memories for simple
word lists.
But this doesn’t mean that all principles of memory apply to all types of
remembering. As we saw in Chapter 7, the rules that govern implicit memory
may be different from those that govern explicit memory. And as we’ve now
seen, some of the factors that play a large role in shaping autobiographical remembering (e.g., the role of emotion) may be irrelevant to other sorts
of memory.
In the end, therefore, our overall theory of memory is going to need
more than one level of description. We’ll need some principles that apply
to only certain types of memory (e.g., principles specifically aimed at emotional remembering). But we’ll also need broader principles, reflecting the
fact that some themes apply to memory of all sorts (e.g., the importance of
memory connections). As the last three chapters have shown, these more
general principles have moved us forward considerably in our understanding of memory in many different domains and have enabled us to illuminate many aspects of learning, of memory retrieval, and of the sources of
memory error.
316 •
C H A P T E R E I G H T Remembering Complex Events
COGNITIVE PSYCHOLOGY AND EDUCATION
remembering for the long term
Sometimes you need to recall things after a short delay — a friend tells you her
address and you drive to her apartment an hour later, or you study for a quiz
that you’ll take tomorrow morning. Sometimes, however, you want to remember things over a much longer time span — perhaps trying to recall things you
learned months or years ago. This longer-term retention is certainly important
in educational settings. Facts that you learn in high school may be crucial for
your professional work later in life. Likewise, facts that you learn in your first
year at college, or in your first year in a job, may be crucial in your third or
fourth year. How, therefore, can we help people to remember things for the
very long term?
The chapter has suggested a two-part answer to this question. First, you’re
more likely to hang on to material that you learned very well in the first
place. The chapter mentions one study in which people tried to recall the
material they’d learned in a college course a decade earlier. In that study, students’ grades in the course were good predictors of how much the students
would remember years after the course was done — and so, apparently, the
better the original learning, the slower the forgetting.
But long-term retention also depends on another factor — whether you
occasionally “revisit” the material you’ve learned. Even a brief refresher
can help enormously. In one study, students were quizzed on little factoids
they had most likely learned at some prior point in their lives (Berger, Hall,
& Bahrick, 1999) — for example, “Who was the first astronaut to walk
on the moon?”; “Who wrote the fable about the fox and the grapes?” In
many cases, the students knew these little facts but couldn’t recall them at
that moment. In that situation, the students were given a quick reminder.
The correct answer was shown to them for 5 seconds, with the simple
instruction that they should look at the answer because they would need
it later on.
Nine days after this reminder, participants were able to recall roughly
half the answers. This obviously wasn’t perfect performance, but it was an
enormous return (an improvement from 0% to 50%) from a very small
investment (5 seconds of “study time”). And it’s likely that a second reminder
a few days later, again lasting just 5 seconds, would have lifted their performance still further and allowed the participants to recall the items after an
even longer delay.
One suggestion, then, is that testing yourself (perhaps with flashcards — with
a cue on one side and an answer on the other) can be quite useful. Flashcards
are often a poor way to learn material, because (as we’ve seen) learning
requires thoughtful and meaningful engagement with the materials you’re
trying to memorize, and running through a stack of flash cards probably
won’t promote that thoughtful engagement. But using flashcards may be an
Cognitive Psychology and Education
•
317
AIDS FOR STUDENTS?
Memory research provides
power­ful lessons for students
hoping to retain what they
are learning in their courses.
excellent way to review material that is already learned — and so a way to
avoid forgetting this material.
Other, more substantial, forms of testing can also be valuable. Think about
what happens each time you take a vocabulary quiz in your Spanish class.
A question like “What’s the Spanish word for ‘bed’?” gives you practice in
retrieving the word, and that practice promotes fluency in retrieval. In addition,
seeing the word (cama) can itself refresh the memory, promoting retention.
The key idea here is the “testing effect.” This term refers to a consistent
pattern in which students who have taken a test have better retention later
on, in comparison to students who didn’t take the initial test. (See, e.g.,
Carpenter, Pashler, & Cepeda, 2009; Glass & Sinha, 2013; Halamish &
Bjork, 2011; Karpicke, 2012; McDermott, Agarwal, D’Antonio, Roediger, &
McDaniel, 2014; Pyc & Rawson, 2012.) This pattern has been documented
with students of various ages (including high school and college students)
and with different sorts of material.
The implications for students should be clear. It really does pay to go
back periodically and review what you’ve learned — including material
you learned earlier this academic year as well as material from previous
years. The review doesn’t have to be lengthy or intense; in the first study
described here, just a 5-second exposure was enough to decrease forgetting
dramatically.
Finally, you shouldn’t complain if a teacher insists on giving frequent
quizzes. Of course, quizzes can be a nuisance, but they serve two functions. First, they can help you assess your learning, so that you can judge
whether — perhaps — you need to adjust your study strategies. Second, the
quizzes actually help you retain what you’ve learned — for days, and probably months, and perhaps even decades after you’ve learned it.
318 •
C H A P T E R E I G H T Remembering Complex Events
For more on this topic . . .
Brown, P. C., Roediger, H. L., & McDaniel, M. A. (2014). Make it stick: The science
of successful learning. New York, NY: Belknap Press.
Putnam, A. L., Nestojko, J. F., & Roediger, H. L. (2016). Improving student learning: Two strategies to make it stick. In J. C. Horvath, J. Lodge, & J. A. C.
Hattie (Eds.), From the laboratory to the classroom: Translating the science
of learning for teachers (pp. 94–121). Oxford, UK: Routledge.
Putnam, A. L., Sungkhasettee, V., & Roediger, H. L. (2016). Optimizing learning
in college: Tips from cognitive psychology. Perspectives on Psychological
Science, 11(5), 652–660.
Cognitive Psychology and Education
•
319
chapter review
SUMMARY
• Memory is usually accurate, but errors do occur
and can be quite significant. In general, these errors
are produced by the connections that link memories to one another and link memories for specific
episodes to other, more general knowledge. These
connections help you because they serve as retrieval
paths. But the connections can also “knit” separate
memories together, making it difficult to keep track
of which elements belong in which memory.
• Some memory errors arise from your understanding of an episode. The understanding promotes
memory for the episode’s gist but also encourages
memory errors. A similar pattern emerges in the
DRM procedure, in which a word related to other
words on a list is (incorrectly) recalled as being part
of the list. Closely related effects arise from schematic knowledge. This knowledge helps you understand an episode, but at the same time a reliance
on schematic knowledge can lead you to remember
an episode as being more “regular,” more “normal,”
than it actually was.
• Memory errors can also arise through the misinformation effect, in which people are exposed to some
(false) suggestion about a previous event. Such suggestions can easily change the details of how an event
is remembered and can, in some cases, plant memories for entire episodes that never occurred at all.
• People seem genuinely unable to distinguish their
accurate memories from their inaccurate ones. This
is because false memories can be recalled with just as
much detail, emotion, and confidence as historically
accurate memories. The absence of a connection
between memory accuracy and memory confidence contrasts with the commonsense belief that
you should rely on someone’s degree of certainty in
assessing their memory. The problem in this commonsense belief lies in the fact that confidence is
influenced by factors (such as feedback) that have no
320
impact on accuracy, and this influence undermines
the linkage between accuracy and confidence.
• While memory errors are easily documented,
cases of accurate remembering can also be observed,
and they are probably more numerous than cases
involving memory error. Memory errors are more
likely, though, in recalling distant events rather than
recent ones. One reason is decay of the relevant
memories; another reason is retrieval failure. Retrieval failure can be either complete or partial; the
tip-of-the-tongue pattern provides a clear example
of partial retrieval failure. Perhaps the most important source of forgetting, though, is interference.
• People have sought various ways of undoing forgetting, including hypnosis and certain drugs. These
approaches, however, seem ineffective. Forgetting can
be diminished, though, through procedures that provide a rich variety of retrieval cues, and it can be avoided
through occasional revisits to the target material.
• Although memory errors are troubling, they may
be the price you pay in order to obtain other advantages. For example, many errors result from the
dense network of connections that link your various
memories. These connections sometimes make it
difficult to recall which elements occurred in which
setting, but the same connections serve as retrieval
paths — and without those connections, you might
have great difficulty in locating your memories in
long-term storage.
• Autobiographical memory is influenced by the
same principles as any other form of memory, but
it is also shaped by its own set of factors. For example, episodes connected to the self are, in general,
better remembered — a pattern known as the “selfreference effect.”
• Autobiographical memories are often emotional,
and this has multiple effects on memory. Emotion
seems to promote memory consolidation, but it may
also produce a pattern of memory narrowing. Some
emotional events give rise to very clear, long-lasting
flashbulb memories. Despite their subjective clarity, these memories can contain errors and in some
cases can be entirely inaccurate. At the extreme
of emotion, trauma has mixed effects on memory.
Some traumatic events are not remembered, but
most traumatic events seem to be remembered for a
long time and in great detail.
• Some events can be recalled even after many
years have passed. In some cases, this is because the
knowledge was learned very well in the first place.
In other cases, occasional rehearsals preserve a
memory for a very long time.
KEY TERMS
intrusion errors (p. 283)
DRM procedure (p. 285)
schema (plural: schemata) (p. 286)
misinformation effect (p. 291)
retention interval (p. 298)
decay theory of forgetting (p. 299)
interference theory (p. 299)
retrieval failure (p. 299)
TOT phenomenon (p. 299)
autobiographical memory (p. 304)
self-schema (p. 304)
consolidation (p. 306)
flashbulb memories (p. 308)
TEST YOURSELF AGAIN
1. What is the evidence that in some circumstances many people will misremember significant events they have experienced?
2. What is the evidence that in some circumstances
people will even misremember recent events?
3. What is the evidence that your understanding
of an episode can produce intrusion errors?
4. What is the DRM procedure, and what results
does this procedure produce?
5. What is schematic knowledge, and what evidence tells us that schematic knowledge can help
us — and also can undermine memory accuracy?
6. What is the misinformation effect? Describe
three different procedures that can produce this
effect.
7. Some people insist that our memories are
consistently accurate in remembering the gist,
or overall content, of an event; when we make
memory errors, they claim, we make mistakes
only about the details within an event. What
evidence allows us to reject this claim?
8. What factors seem to undermine the relationship
between your degree of certainty in a memory
and the likelihood that the memory is accurate?
9. Explain the mechanisms hypothesized by each
of the three major theories of forgetting: decay,
interference, and retrieval failure.
10. What techniques or procedures seem ineffective as a means of “un-doing” forgetting? What
techniques or procedures seem to diminish or
avoid forgetting?
321
11. Explain why the mechanisms that produce
memory errors may actually be mechanisms
that help us in important ways.
13. What is a flashbulb memory? Are flashbulb
memories distinctive in how accurate they seem
to be?
12. What is memory consolidation?
THINK ABOUT IT
1.People sometimes compare the human eye
to a camera, and compare human memory
to a video recorder (like TiVo or the video
recorder on your smartphone). Ironically,
though, there are important ways in which
your memory is worse than a video recorder,
and also important ways in which it’s far better
than a video recorder. Describe both versions
of this comparison — the ways in which video
recorders are superior, and the ways in which
your memory is superior.
E eBOOK DEMONSTRATIONS & ESSAYS
Go to http://digital.wwnorton.com/cognition7 for the online demonstrations and essays
relevant to this chapter. These can be found in the ebook.
Online Demonstrations
Online Applying Cognitive Psychology and the
Law Essays
• Demonstration 8.1: Associations and Memory
• Cognitive Psychology and the Law: Jurors’
Error
Memory
• Demonstration 8.2: Memory Accuracy and
Confidence
• D
emonstration 8.3: The Tip-of-the-Tongue Effect
• Demonstration 8.4: Childhood Amnesia
COGNITION LABS
Go to http://digital.wwnorton.com/cognition7 to experience interactive online psychology
labs relevant to this chapter.
322
Knowledge
4
part
I
n Parts 2 and 3, we saw case after case in which your interactions with
the world are guided by knowledge. In perceiving, for example, you make
inferences guided by knowledge about the world’s regular patterns. In
attending, you anticipate inputs guided by your knowledge about what’s likely
to occur. In learning, you connect new information to things you already know.
But what is knowledge? How is it represented in your mind? How do you locate
knowledge in memory when you need it? We’ve already taken some steps toward
answering these questions — by arguing that knowledge is represented in the mind
by means of a network of interconnected nodes. In this section, we’ll expand this
proposal in important ways. In Chapter 9, we’ll describe the basic building blocks
of knowledge — individual concepts — and consider several hypotheses about how
concepts are represented in the mind. Because each hypothesis captures a part of
the truth, we’ll be driven toward a several-part theory that combines the various
views. We’ll also see that knowledge about individual concepts depends on linkages
to other, related concepts. For example, you can’t know what a “dog” is without also
understanding what an “animal” is, what a “living thing” is, and so on. As a result,
connections among ideas will be crucial here, just as they were in previous chapters.
Chapters 10 and 11 then focus on two special types of knowledge: knowledge about language and knowledge about visual images. In Chapter 10, we’ll
see that your knowledge of language is highly creative, allowing you to produce
new words and new sentences that no one has ever used before. But at the
same time, the creativity is constrained, so there are some words and sequences
of words that are considered unacceptable by virtually any language-user. In
order to understand this pattern of “constrained creativity,” we’ll consider the
possibility that language knowledge involves abstract rules that are, in some
way, honored by every user of the language.
In Chapter 11, we’ll see that mental images involve representations that are
distinct from those involved in other forms of knowledge, but we’ll also consider some of the ways in which memory for visual appearances is governed by
the same principles as other forms of knowledge.
323
9
chapter
Concepts and
Generic Knowledge
what if…
In Chapter 8, we mentioned people who have superior autobiographical recall. It’s remarkable how much
these individuals can remember — but some people, it turns out, remember
even more. One might say that these people have “perfect memories,”
but this terminology would be misleading.
We begin with a work of fiction. In a wonderful short story titled
“Funes the Memorious,” the Argentine writer Jorge Luis Borges
describes a character — Funes — who never forgets anything. But rather
than being proud of this capacity, Funes is immensely distressed by his
memorial prowess: “My memory, sir, is like a garbage heap” (p. 152).
Among other problems, Funes complains that he’s incapable of thinking in general terms. He remembers so much about how individuals differ
that he has a hard time focusing on what they might have in common:
“Not only was it difficult for him to comprehend that the generic symbol
dog embraces so many unlike individuals of diverse size and form; it bothered him that the dog at 3:14 (seen from the side) should have the same
name as the dog at 3:15 (seen from the front)” (Borges, 1964, p. 153).
Funes is a fictional character, but consider the actual case of Solomon
Shereshevsky (Luria, 1968). Shereshevsky, like Funes, never forgot anything. After hearing a lengthy speech, he could repeat it back word for
word. If shown a complex mathematical formula (even one that had
no meaning for him), he could reproduce it perfectly months later. He
effortlessly memorized poems written in languages he didn’t understand. And Shereshevsky’s flawless retention wasn’t the result of some
deliberate trick or strategy. Just the opposite: Shereshevsky seemed to
have no choice about his level of recall.
Like Funes, Shereshevsky wasn’t well served by his extraordinary
memory. He was so alert to the literal form of his experiences that he
couldn’t remember their deeper implications. Similarly, he had difficulty
recognizing faces because he was so alert to the changes in a face from
one view to the next. And, like Funes, Shereshevsky was often distracted
by the detail of his own recollections, so he found it difficult to think in
abstract terms.
There are, of course, settings in which you do want to remember the
specific episodes of your life. You want to recall what you saw at a crime
scene, holding to the side (as best you can) the information you picked
up later in your conversation with the police. You hope to remember
325
preview of chapter themes
•
asic concepts — like “chair” and “dog” — are the building
B
blocks of all knowledge. However, attempts at defining
these concepts usually fail because we easily find exceptions to any definition that might be proposed.
•
hese beliefs may be represented in the mind as proposiT
tions encoded in a network structure. Alternatively, they
may be represented in a distributed form in a connectionist network.
•
his leads to a suggestion that knowledge of these conT
cepts is cast in terms of probabilities — so that a creature
that has wings and feathers, and that flies and lays eggs, is
probably a bird.
•
•
any results are consistent with this probabilistic idea and
M
show that the more a test case resembles the “prototype”
for a category, the more likely people are to judge the case
as being in that category.
e are driven, therefore, to a multipart theory of conW
cepts. Your conceptual knowledge likely includes a
prototype for each category and also a set of remembered exemplars. But you also seem to have a broad set
of beliefs about each concept — beliefs that provide a
“theory” for why the concept takes the form it does, and
you use this theory in a wide range of judgments about
the concept.
•
ther results, however, indicate that conceptual knowlO
edge includes other beliefs — beliefs that link a concept to
other concepts and also specify why the concept is as it is.
what you read in your textbook, trying to ignore the (possibly bogus)
information you heard from your roommate. Funes and Shereshevsky
obviously excel in this type of memory, but their limitations remind us
that there are also disadvantages for this type of particularized recall. In
many settings, you want to set aside the details of this or that episode
and, instead, weave your experiences together so that you can pool
information received from various sources. This allows you to create a
more complete, more integrated type of knowledge — one that allows
you to think about dogs in general rather than focusing on this view of
that dog; or one that helps you remember what your friend’s face generally looks like rather than what she looked like, say, yesterday at 1:42 in
the afternoon. This more general type of knowledge is surely drawn from
your day-to-day experience, but it is somehow abstracted away from
that experience. What is this more general type of knowledge?
Understanding Concepts
Imagine that a friend approaches you and boasts that he knows what a
“spoon” is or what a “shoe” is. You’d probably be impressed by your
friend’s foolishness, not by his knowledge. After all, concepts like these are so
ordinary, so straightforward, that there seems to be nothing special about
knowing — and being able to think about — these simple ideas.
However, ordinary concepts like these are the building blocks out of
which all knowledge is created, and as we’ve seen in previous chapters, you
depend on your knowledge in many aspects of day-to-day functioning. Thus,
you know what to pay attention to in a restaurant because you understand
326 •
C H A P T E R N I N E Concepts and Generic Knowledge
the basic concept of “restaurant.” You’re able to understand a simple story
about a child checking her piggy bank because you understand the concepts
of “money,” “shopping,” and so on.
The idea, then, is that you need concepts in order to have knowledge, and
you need knowledge in order to function. In this way, your understanding of
ideas like “spoon” and “shoe” might seem commonplace, but it is an ingredient without which cognition cannot proceed.
But what exactly does it mean to understand concepts like these? How
is this knowledge represented in the mind? In this chapter, we’ll begin with
the hypothesis that understanding a concept is like knowing a dictionary
definition — and so, if someone knows what a “house” is, or a “taxi,” he or
she can offer something like a definition for these terms — and likewise for
all the other concepts in each person’s knowledge base. As we’ll see, though,
this hypothesis quickly runs into problems, so we’ll need to turn to a more
complicated proposal.
Definitions: What Is a “Dog”?
You know perfectly well what a dog is. But what is it that you know? One possibility is that your knowledge is somehow akin to a dictionary definition —
that is, what you know is something like: “A dog is a creature that (a) is an
animal, (b) has four legs, (c) barks, (d) wags its tail.” You could then use this
definition in straightforward ways: When asked whether a candidate creature
is a dog, you could use the definition as a checklist, scrutinizing the candidate
for the various defining features. When told that “a dog is an animal,” you
would know that you hadn’t learned anything new, because this information
is already contained within the definition. If you were asked what dogs, cats,
and horses have in common, you could scan your definition of each one looking for common elements.
This proposal is correct in some cases, and so, for example, you certainly
know definitions for concepts like “triangle” and “even number.” But what about
more commonplace concepts? The concern here was brought to light by the
20th-century philosopher Ludwig Wittgenstein, who argued (e.g., Wittgenstein,
1953) that the simple terms we all use every day actually don’t have definitions.
For example, consider the word “game.” You know this word and can use it
sensibly, but what is a game? As an approach to this question, we could ask,
for example, about the game of hide-and-seek. What makes hide-and-seek a
“game”? Hide-and-seek (a) is an activity most often practiced by children, (b) is
engaged in for fun, (c) has certain rules, (d) involves several people, (e) is in some
ways competitive, and (f) is played during periods of leisure. All these are plausible attributes of games, and so we seem well on our way to defining “game.”
But are these attributes really part of the definition of “game”? What about the
Olympic Games? The competitors in these games aren’t children, and runners in
marathon races don’t look like they’re having a lot of fun. Likewise, what about
card games played by one person? These are played alone, without competition.
For that matter, what about the case of professional golfers?
Understanding Concepts
•
327
THE HUNT FOR DEFINITIONS
It is remarkably difficult to define even very familiar terms. For example, what is a “dog”?
Most people include “has fur” in the definition, but what about the hairless Chihuahua?
Many people include “communicates by barking” in the definition, but what about the
Basenji (one of which is shown here) — a breed of dog that doesn’t bark?
TEST YOURSELF
1.Consider the word
“chair.” Name some
attributes that might
plausibly be included
in a definition for this
word. But, then, can
you describe objects
that you would count
as chairs even though
they don’t have one
or more of these
attributes?
2.What does it mean to
say that there is a “family
resemblance” among the
various animals that we
call “dogs”?
328 •
It seems that for each clause of the definition, we can find an exception —
an activity that we call a “game” but that doesn’t have the relevant characteristic. And the same is true for almost any concept. We might define “shoe” as
an item of apparel made out of leather, designed to be worn on the foot.
But what about wooden shoes? What about a shoe designed by a master
shoemaker, intended only for display? What about a shoe filled with cement,
which therefore can’t be worn? Similarly, we might define “dog” in a way
that includes four-leggedness, but what about a dog that has lost a limb in
some accident? We might specify “communicates by barking” as part of the
definition of dog, but what about the African Basenji, which has no bark?
Family Resemblance
It seems, then, we can’t say things like “A dog is a creature that has fur and
four legs and barks.” That’s because we easily find exceptions to this rule
(a hairless Chihuahua; a three-legged dog; the barkless Basenji). But surely
we can say, “Dogs usually are creatures that have fur, four legs, and bark, and
a creature without these features is unlikely to be a dog.” This probabilistic
phrasing preserves what’s good about definitions — the fact that they do name
relevant features, shared by most members of the category. But this phrasing
also allows a degree of uncertainty, some number of exceptions to the rule.
C H A P T E R N I N E Concepts and Generic Knowledge
In a similar spirit, Wittgenstein proposed that members of a category have
a family resemblance to one another. To understand this term, think about
an actual family — your own, perhaps. There are probably no “defining
features” for your family — features that every family member has. Nonetheless, there are features that are common in the family, and so, if we consider
family members two or three at a time, we can usually find shared attributes.
For example, you, your brother, and your mother might all have the family’s
beautiful red hair and the same wide lips; as a result, you three look alike to
some extent. Your sister, however, doesn’t have these features. But she’s still
recognizable as a member of the family because (like you and your father)
she has the family’s typical eye shape and the family’s distinctive chin. In this
way, the common features in the family depend on what “subgroup” you’re
considering — hair color shared for these family members; eye shape shared
by those family members; and so on.
One way to think about this pattern is by imagining the “ideal” for each
family — someone who has all of the family’s features. (In our example,
this would be a wide-lipped redhead with the right eye and chin shapes.) In
many families, this person may not exist, so perhaps there’s nobody who has
every one of the family’s distinctive features — and so no one who looks like
the “perfect Jones” (or the “perfect Martinez” or the “perfect Goldberg”).
Nonetheless, each member of the family shares at least some features with
this ideal — and therefore has some features in common with other family members. This feature overlap is why the family members resemble one
another, and it’s how we manage to recognize these individuals as all belonging to the same family.
Wittgenstein proposed that ordinary categories like “dog” or “game” or
“furniture” work in the same way. There may be no features that are shared
by all dogs or all games. Even so, we can identify “characteristic features” for
each category — features that many (perhaps most) category members have.
These are the features that enable you to recognize that a dog is a dog, a game
is a game, and so on.
There are several ways we might translate all these points into a psychological theory, but one influential translation was proposed by psychologist Eleanor Rosch in the mid-1970s (Rosch, 1973, 1978; Rosch & Mervis,
1975; Rosch, Mervis, Gray, Johnson, & Boyes-Braem, 1976). Let’s look at
her model.
TYPICALITY IN
FAMILIES
In the Smith family, many (but
not all) of the brothers have
dark hair, so dark hair is typical for the family (i.e., is found
in many family members)
but doesn’t define the family
(i.e., is not found in all family
members). Likewise, wearing
glasses is typical for the family but not a defining feature;
so is having a mustache and a
big nose. Many concepts have
the same character — with
many features shared among
the instances of the concept,
but no features shared by all
of the instances.
Prototypes and Typicality Effects
One way to think about definitions is that they set the “boundaries” for a
category. If a test case has certain attributes, then it’s “inside” the boundaries. If a test case doesn’t have the defining attributes, then it’s “outside” the
category. Prototype theory, in contrast, begins with a different tactic: Perhaps
the best way to identify a category is to specify the “center” of the category,
rather than the boundaries. Just as we spoke earlier about the “ideal” family
member, perhaps the concept of “dog” is represented in the mind by some
Prototypes and Typicality Effects
•
329
depiction of the “ideal” dog, and all judgments about dogs are made with
reference to this ideal. Likewise for “bird” or “house” or any other concept
in your repertoire — in each case, the concept is represented by the appropriate prototype.
In most cases, this “ideal” — the prototype — will be an average of the
various category members you’ve encountered. So, for example, the prototype dog will be the average color of the dogs you’ve seen, the average size
of the dogs you’ve seen, and so forth. (Notice, then, that different people,
each with their own experiences, will have slightly different prototypes.)
No matter what the specifics of the prototype, though, you’ll use this
“ideal” as the benchmark for your conceptual knowledge. Thus, whenever
you use your conceptual knowledge, your reasoning is done with reference
to the prototype.
Prototypes and Graded Membership
To make these ideas concrete, imagine that you’re trying to decide whether
a creature currently before your eyes is or is not a dog. In making this
decision, you’ll compare the creature with the prototype in your memory.
If there’s no similarity, the creature standing before you is probably not
in the category; if there’s considerable similarity, you draw the opposite
conclusion.
This sounds plausible enough, but note an important implication.
Membership in a category depends on resemblance to the prototype, and
resemblance is a matter of degree. (After all, some dogs are likely to resemble
the prototype closely, while others will have less in common with this ideal.)
As a result, membership in the category isn’t a simple “yes or no” decision;
CATEGORIES HAVE PROTOTYPES
As the text describes, people seem to have a prototype
in their minds for a category like “dog.” For many people,
the German shepherd shown here is close to that prototype, and the other dogs depicted are more distant from
the prototype.
330 •
C H A P T E R N I N E Concepts and Generic Knowledge
instead, it’s a matter of “more” or “less.” In technical terms, we’d say that
categories, on this view, have a graded membership, such that objects closer
to the prototype are “better” members of the category than objects farther
from the prototype. Basically, the idea is that some dogs are “doggier” than
others, some books “bookier” than others, and so on for all the other categories you can think of.
Testing the Prototype Notion
This proposal — that mental categories have a graded membership — was
tested in a series of experiments conducted years ago. For example, in
classic studies using a sentence verification task, research participants were
presented with a series of sentences, and their job was to indicate (by pressing the appropriate button) whether each sentence was true or false. In this
procedure, participants’ responses were slower for sentences like “A penguin
is a bird” than for sentences like “A robin is a bird”; slower for “An Afghan
hound is a dog” than for “A German shepherd is a dog” (Smith, Rips, &
Shoben, 1974).
Why should this be? According to a prototype perspective, participants
chose their response (“true” or “false”) by comparing the thing mentioned
(e.g., penguin) to their prototype for that category (i.e., their bird prototype).
When there was close similarity between the test case and the prototype,
participants could make their decisions quickly; in contrast, judgments about
items distant from the prototype took more time. And given the results, it
seems that penguins and Afghans are more distant from their respective prototypes than are robins and German shepherds.
Other early results can also be understood in these terms. For example, in a production task we simply ask people to name as many birds or
dogs as they can. According to a prototype view, they’ll do this task by first
locating their bird or dog prototype in memory and then asking themselves
what resembles this prototype. In essence, they’ll start with the center of the
category (the prototype) and work their way outward from there. So birds
close to the prototype should be mentioned first; birds farther from the prototype, later on.
By this logic, the first birds mentioned in the production task should
be the birds that yielded fast response times in the verification task; that’s
because what matters in both tasks is proximity to the prototype. Likewise, the birds mentioned later in production should have yielded slower
response times in verification. This is exactly what happened (Mervis,
Catlin, & Rosch, 1976).
In fact, this outcome sets the pattern of evidence for prototype theory.
Over and over, in category after category, members of a category that are
“privileged” on one task (e.g., they yield the fastest response times) turn out
also to be privileged on other tasks (e.g., they’re most likely to be mentioned).
As another illustration of this pattern, consider the data from rating tasks. In
these tasks, participants are given instructions like these: “We all know that
Prototypes and Typicality Effects
•
331
some birds are ‘birdier’ than others, some dogs are ‘doggier’ than others, and
so on. I’m going to present you with a list of birds or of dogs, and I want you
to rate each one on the basis of how ‘birdy’ or ‘doggy’ it is” (Rosch, 1975;
also Malt & Smith, 1984).
People are easily able to make these judgments, and quite consistently they
rate items as being very “birdy” or “doggy” when these instances are close to
the prototype (as determined in the other tasks). They rate items as being less
“birdy” or “doggy” when these are farther from the prototype. This finding
suggests that once again, people perform the task by comparing the test item
to the prototype (see Table 9.1).
Basic-Level Categories
It does seem, then, that certain category members are “privileged,” just as the
prototype theory proposes. It turns out, also, that certain types of category are
privileged — in their structure and the way they’re used. For example, imagine
TABLE 9.1
ARTICIPANTS’ TYPICALITY RATINGS FOR THE
P
CATEGORY “FRUIT” AND THE CATEGORY “BIRD”
Fruit
Rating
Bird
Rating
Apple
6.25
Robin
6.89
Peach
5.81
Bluebird
6.42
Pear
5.25
Seagull
6.26
Grape
5.13
Swallow
6.16
Strawberry
5.00
Falcon
5.74
Lemon
4.86
Mockingbird
5.47
Blueberry
4.56
Starling
5.16
Watermelon
4.06
Owl
5.00
Raisin
3.75
Vulture
4.84
Fig
3.38
Sandpiper
4.47
Coconut
3.06
Chicken
3.95
Pomegranate
2.50
Flamingo
3.37
Avocado
2.38
Albatross
3.32
Pumpkin
2.31
Penguin
2.63
Olive
2.25
Bat
1.53
Ratings were made on a 7-point scale, with 7 corresponding to the highest typicality.
Note also that the least “birdy” of the birds isn’t (technically speaking) a bird at all!
( after malt & smith , 1984)
332 •
C H A P T E R N I N E Concepts and Generic Knowledge
FIGURE 9.1 BASIC VERSUS
SUPERORDINATE LABELING
What is this? The odds are good that
you would answer, “It’s a chair,” using
the basic-level description rather than
the more general label (“It’s a piece of
furniture”) or a more specific description
(“It’s an upholstered armchair”) — even
though these other descriptions would
certainly be correct.
that we show you a picture like the one in Figure 9.1 and ask, “What is this?”
You’re likely to say “a chair” and unlikely to offer a more specific response
(“upholstered armchair”) or a more general one (“an item of furniture”).
Likewise, we might ask, “How do people get to work?” In responding, you’re
unlikely to say, “Some people drive Fords; some drive Toyotas.” Instead, your
answer is likely to use more general terms, such as “cars,” “trains,” and “buses.”
In keeping with these observations, Rosch and others have argued that
there is a “natural” level of categorization, neither too specific nor too general,
that people tend to use in their conversations and their reasoning. The special
status of this basic-level categorization can be demonstrated in many ways.
Basic-level categories are usually represented in our language via a single
word, while more specific categories are identified with a phrase. Thus,
“chair” is a basic-level category, and so is “apple.” The more specific (subordinate) categories of “lawn chair” or “kitchen chair” aren’t basic level; neither is
“Granny Smith apple” or “Golden Delicious apple.”
We’ve already suggested that if you’re asked to describe an object, you’re
likely to use the basic-level term. In addition, if asked to explain what members of a category have in common with one another, you have an easy time
with basic-level categories (“What do all chairs have in common?”) but some
difficulty with more inclusive (superordinate) categories (“What does all furniture have in common?”). Moreover, children learning to talk often acquire
basic-level terms earlier than either the more specific subcategories or the more
general, more encompassing categories. In these (and other) ways, basic-level
categories do seem to reflect a natural way to categorize the objects in our
world. (For more on basic-level categories, see Corter & Gluck, 1992; Murphy,
2016; Pansky & Koriat, 2004; Rogers & Patterson, 2007; Rosch et al., 1976.)
TEST YOURSELF
3.Why is graded membership a consequence
of representing the
category in terms of a
prototype?
4.What tasks show us that
concept judgments often
rely on prototypes and
typicality?
5.Give an example of a
basic-level category,
and then name some
of the subcategories
within this basic-level
grouping.
Prototypes and Typicality Effects
•
333
Exemplars
Let’s return, though, to our main agenda. As we’ve seen, a broad range of
tasks reflects the graded membership of mental categories. In other words,
some members of the categories are “better” than others, and the better members are recognized more readily, mentioned more often, judged to be more
typical, and so on. (For yet another way you’re influenced by typicality, see
Figure 9.2.) All of this fits well with the idea that conceptual knowledge is
represented via a prototype and that we categorize by making comparisons
to that prototype. It turns out, though, that your knowledge about “birds”
and “fruits” and “shoes” and so on also includes another element.
FIGURE 9.2
TYPICALITY AND ATTRACTIVENESS
Typicality influences many judgments about category members, including attractiveness. Which of these pictures
shows the most attractive-looking fish? Which one shows the least attractive-looking? In several studies, participants’ ratings of attractiveness have been closely related to (other participants’) ratings of typicality — so that
people seem to find more-typical category members to be more attractive (e.g., Halberstadt & Rhodes, 2003).
334 •
C H A P T E R N I N E Concepts and Generic Knowledge
Analogies from Remembered Exemplars
Imagine that we place a wooden object in front of you and ask, “Is this
a chair?” According to the prototype view, you’ll answer by calling up
your chair prototype from memory and then comparing the candidate
to that prototype. If the resemblance is great, you’ll announce, “Yes, this
is a chair.”
But you might make this decision in a different way. You might notice that
the object is very similar to an object in your Uncle Jerry’s living room, and
you know that Uncle Jerry’s object is a chair. (After all, you’ve seen Uncle
Jerry sitting in the thing, reading his newspaper; you’ve heard Jerry referring
to the thing as “my chair,” and so on.) These points allow an easy inference:
If the new object resembles Jerry’s, and if Jerry’s object is a chair, then it’s a
safe bet that the new object is a chair too.
The idea here is that in some cases categorization relies on knowledge
about specific category members (e.g., “Jerry’s chair”) rather than the prototype (e.g., the ideal chair). This process is referred to as exemplar-based
reasoning, with an exemplar defined as a specific remembered instance — in
essence, an example.
The exemplar-based approach is in many ways similar to the prototype view. According to each of these proposals, you categorize objects by
comparing them to a mentally represented “standard.” The difference
between the views lies in what that standard is. For prototype theory,
the standard is the prototype — an average representing the entire category; for exemplar theory, the standard is provided by whatever example of the category comes to mind (and different examples may come to
mind on different occasions). In either case, the process is then the same.
You assess the similarity between a candidate object and the standard. If
the resemblance is great, you judge the candidate as being within the
relevant category; if the resemblance is minimal, you seek some alternative
categorization.
A Combination of Exemplars and Prototypes
There is, in fact, reason for you to rely on prototypes and on exemplars in
your thinking about categories. Prototypes provide an economical representation of what’s typical for a category, and there are many circumstances in
which this quick summary is useful. But exemplars, for their part, provide
information that’s lost from the prototype — including information about the
variability within the category.
To see how this matters, consider the fact that people routinely “tune”
their concepts to match the circumstances. For example, they think about
birds differently when considering Chinese birds than when thinking about
American birds; they think about gifts differently when considering gifts for
a student rather than gifts for a faculty member (Barsalou, 1988; Barsalou
& Sewell, 1985). In fact, people can adjust their categories in fairly precise
Exemplars
•
335
ways: not just “gift,” but “gift for a 4-year-old” or “gift for a 4-year-old
who recently broke her wrist” or “gift for a 4-year-old who likes sports but
recently broke her wrist.” This pliability in concepts is easy to understand if
people are relying on exemplars; after all, different settings, or different perspectives, would trigger different memories and so bring different exemplars
to mind.
It’s useful, then, that conceptual knowledge includes prototypes and
exemplars, because each has its own advantages. However, the mix of
exemplar and prototype knowledge may vary from person to person and
from concept to concept. One person might have extensive knowledge about
individual horses, so she has many exemplars in memory; the same person
might have only general information (a prototype, perhaps) about snowmobiles. Some other person might show the reverse pattern. And for all people,
the pattern of knowledge might depend on the size of the category and how
easily confused the category memories are with one another — with exemplars
being used when the individuals are more distinct. (For further discussion,
see Murphy, 2016; Rips, Smith, & Medin, 2012; Rouder & Ratcliff, 2006;
Smith, Zakrzewski, Johnson, & Valleau, 2016; Vanpaemel & Storms, 2008.
For discussion of the likely neural basis for exemplar storage, see Ashby &
Rosedahl, 2017; also see Figure 9.3.)
Overall, though, it cannot be surprising that people can draw on either
prototypes or exemplars when thinking about concepts. The reason is that
the two types of information are used in essentially the same way. In either
case, an object before your eyes triggers some representation in memory
(either a representation of a specific instance, according to exemplar theory,
or the prototype, according to prototype theory). In either case, you assess the
FIGURE 9.3 DISTINCTIONS
WITHIN CATEGORIES
The chapter suggests that you have
knowledge of both exemplars and
prototypes. As a further complication, you also have special knowledge
about distinctive individuals within a
category. Thus, you know that Kermit
has many frogly properties (he’s
green, he eats flies, he hops) but also
has unusual properties that make him
a rather unusual frog (since, after all,
he can talk, he can sing, and he’s in
love with a pig).
336 •
C H A P T E R N I N E Concepts and Generic Knowledge
resemblance between this conceptual knowledge, supplied by memory, and
the novel object before you: “Does this object resemble my sister’s couch?” If
so, the object is a couch. “Does the object resemble my prototype for a soup
bowl?” If so, it’s probably a soup bowl.
Given these similarities, it seems plausible that we might merge the proto­
type and exemplar proposals, with each of us on any particular occasion
relying on whichever sort of information (exemplar or prototype) comes to
mind more readily.
The Difficulties with Categorizing
via Resemblance
TEST YOURSELF
6.What is similar in the
processes of categorizing via a prototype
and the processes of
categorizing via an
exemplar? What is
different between
these two types of
processes?
We’re moving, it seems, toward a clear-cut set of claims. First, for most concepts, definitions are not available. Second, for many purposes, you don’t
need a definition and can rely instead on a mix of prototypes and exemplars.
Third, typicality — the degree to which a particular object or situation or
event is typical for its kind — plays a large role in people’s thinking, with
more-typical category members being “privileged” in many ways. Fourth,
typicality is exactly what we would expect if category knowledge does, in
fact, hinge on prototypes and exemplars.
This reasoning seems straightforward enough. However, some results
don’t fit into this picture, so the time has come to broaden our conception of
concepts.
The Differences between Typicality
and Categorization
If you decide that Mike is a bully, or that an event is a tragedy, or that a particular plant is a weed, it’s usually because you’ve compared the “candidate”
in each case to the relevant prototype or exemplar. In essence, you’ve asked
yourself how much the candidate person (or event, or object) resembles a
typical member of the target category. If the resemblance is strong, you decide
the candidate is typical for the category, and likely a member of that category.
If the resemblance is poor, you decide the candidate isn’t at all typical, and its
category status is (at best) uncertain.
The essential point, then, is that judgments of category membership
depend on judgments of typicality, and so these two types of judgment will
inevitably go hand in hand. This point certainly fits with the data we’ve
seen so far, but it doesn’t fit with some other results — results that show
no linkage at all between judgments of category membership and judgments
of typicality.
Armstrong, Gleitman, and Gleitman (1983) gave participants this peculiar
instruction: “We all know that some numbers are even-er than others. What
I want you to do is to rate each of the numbers on this list for how good an
example it is for the category ‘even number.’” Participants were then given a
The Difficulties with Categorizing via Resemblance
•
337
list of numbers (4, 16, 32, and so on) and had to rate “how even” each number was. The participants thought this was a strange task but followed the
instruction nonetheless — and, interestingly, were quite consistent with one
another in their judgments (see Table 9.2).
Of course, participants responded differently (and correctly!) if asked in a
direct way which numbers on the list were even and which were not. Apparently, then, participants could judge category membership as easily as they
could judge typicality, but — importantly — these judgments were entirely
independent of each other. Thus, for example, participants responded that
4 is a more typical even number than 7,534, but they knew this has nothing
to do with the fact that both are unmistakably in the category “even number.”
Clearly, therefore, there’s some basis for judging category membership that’s
separate from the assessment of typicality.
One might argue, though, that mathematical concepts like “even number”
are somehow special, and so their status doesn’t tell us much about other,
more “ordinary” concepts. However, this suggestion is quickly rebutted,
because other concepts show a similar distinction between category membership and typicality. For example, robins strike us as being closer to the typical
bird than penguins are; even so, most of us are certain that both robins and
penguins are birds. Likewise, Moby Dick was definitely not a typical whale,
but he certainly was a whale; Abraham Lincoln wasn’t a typical American,
but he was an American. These informal observations, like the even-number
result, drive a wedge between typicality and category membership — a wedge
that doesn’t fit with our theory so far.
TABLE 9.2
ARTICIPANTS’ TYPICALITY RATINGS FOR
P
WELL-DEFINED CATEGORIES
EVEN NUMBER
Stimulus
ODD NUMBER
Typicality Rating
Stimulus
Typicality Rating
4
5.9
3
5.4
8
5.5
7
5.1
10
5.3
23
4.6
18
4.4
57
4.4
34
3.6
501
3.5
106
3.1
447
3.3
Participants rated numbers on how typical they were for the category “even number.”
Ratings were on a scale from 0 to 7, with 7 meaning the item was (in the participants’
view) very typical. Mathematically this is absurd: Either a number is even (divisible
by 2 without a remainder) or it is not. Even so, participants rated some numbers
as “evener” than others, and likewise rated some odd numbers as being “odder”
than others.
( after armstrong et al ., 1983)
338 •
C H A P T E R N I N E Concepts and Generic Knowledge
How are category judgments made when they don’t rely on typicality?
As an approach to this question, let’s think through an example. Consider a
lemon. Paint the lemon with red and white stripes. Is it still a lemon? Most
people say that it is. Now, inject the lemon with sugar water, so it has a sweet
taste. Then, run over the lemon with a truck, so that it’s flat as a pancake.
What have we got at this point? Do we have a striped, artificially sweet, flattened lemon? Or do we have a non-lemon? Most people still accept this poor,
abused fruit as a lemon, but consider what this judgment involves. We’ve
taken steps to make this object more and more distant from the prototype
and also very different from any specific lemon you’ve ever encountered (and
therefore very different from any remembered exemplars). But this seems
not to shake your faith that the object remains a lemon. To be sure, we have
a not-easily-recognized lemon, an exceptional lemon, but it’s still a lemon.
Apparently, something can be a lemon with virtually no resemblance to
other lemons.
Related points emerge in research with children. In one early study,
preschool children were asked what makes something a “coffeepot,” a “raccoon,” and so on (Keil, 1986). As a way of probing their beliefs, the children
were asked whether it would be possible to turn a toaster into a coffeepot.
Children realized that we’d have to widen the holes in the top of the toaster
CATEGORIZATION OUTSIDE
OF TYPICALITY
Moby Dick was not a typical whale, but he unmistakably
was a whale. Clearly, then, in some settings, typicality
can be separated from category membership.
The Difficulties with Categorizing via Resemblance
•
339
and fix things so that the water wouldn’t leak out of the bottom. We’d also
need to design a place to put the coffee grounds. But the children saw no
obstacles to these manipulations and were quite certain that with these
adjustments in place, we would have created a bona fide coffeepot.
Things were different, though, when the children were asked a parallel
question — whether one could, with suitable adjustments, turn a skunk into
a raccoon. The children understood that we could dye the skunk’s fur, teach
it to climb trees, and, in general, teach it to behave in a raccoon-like fashion.
Even with these adjustments, the children steadfastly denied that we would
have created a raccoon. A skunk that looks, sounds, and acts just like a raccoon might be a very peculiar skunk, but it would be a skunk nonetheless.
(For other evidence suggesting that people reason differently about naturally
occurring items like raccoons and manufactured items like coffeepots, see
Caramazza & Shelton, 1998; Estes, 2003; German & Barrett, 2005; Levin,
Takarae, Miner, & Keil, 2001. Also see “Different Profiles for Different
Concepts” section on p. 347.)
What lies behind all these judgments? If people are asked why the abused
lemon still counts as a lemon, they’re likely to mention that it grew on a
lemon tree, is genetically a lemon, and is still made up of (mostly) the “right
stuff.” It’s these “deep” features that matter, not the lemon’s current properties. And so, too, for raccoons: In the children’s view, being a raccoon
isn’t merely a function of having the relevant features; instead, according
to the children, the key to being a raccoon involves (among other things)
having a raccoon mommy and a raccoon daddy. In this way, a raccoon,
just like a lemon, is defined in ways that refer to deep properties and not to
mere appearances.
Notice, though, that these claims about an object’s deep properties
depend on a web of other beliefs — beliefs that are, in each case, “tuned”
to the category being considered. Thus, you’re more likely to think that a
creature is a raccoon if you’re told that it has raccoons as parents, but this is
true only because you have some ideas about how a creature comes to be a
raccoon — ideas that are linked to your broader understanding of biological
categories and inheritance. It’s this understanding that tells you that parentage is relevant here. If this point isn’t clear, consider as a contrasting case the
steps you’d go through in deciding whether Judy really is a doctor. In this
case, you’re unlikely to worry about whether Judy has a doctor mommy and
a doctor daddy, because your beliefs tell you, of course, that for this category
parentage doesn’t matter.
As a different example, think about the category “counterfeit money.” A
counterfeit bill, if skillfully produced, will have a nearly perfect resemblance
to the prototype for legitimate money. Despite this resemblance, you understand that a counterfeit bill isn’t in the category of legitimate money, so here,
too, your categorization doesn’t depend on typicality. Instead, your categorization depends on a web of other beliefs, including beliefs about circumstances
of printing. A $20 bill is legitimate, you believe, only if it was printed
with the approval of, and under the supervision of, the relevant government
340 •
C H A P T E R N I N E Concepts and Generic Knowledge
A SPANIEL, NOT A WOLF,
IN SHEEP’S CLOTHING?
Both of these creatures resemble the prototype for sheep,
and both resemble many sheep
exemplars you’ve seen (or perhaps read about). But are they
really sheep?
agencies. And once again, these beliefs arise only because you have a broader
understanding of what money is and how government regulations apply to
monetary systems. In other words, you consider circumstances of printing
only because your understanding tells you that the circumstances are relevant
here, and you won’t consider circumstances of printing in a wide range of
other cases. If asked, for example, whether a copy of the Lord’s Prayer is
“counterfeit,” your beliefs tell you that the Lord’s Prayer is the Lord’s Prayer
no matter where (or by whom) it was printed. Instead, what’s crucial for the
prayer’s “authenticity” is simply whether the words are the correct words.
The Complexity of Similarity
Let’s pause to review. There’s no question that judgments about categories
are often influenced by typicality, and we’ll need to account for this fact in
our theorizing. Sometimes, though, category judgments are independent of
typicality: You judge some candidates to be category members even though
they don’t resemble the prototype (think about Moby Dick or the abused
lemon). You judge some candidates not to be in the category even though
they do resemble the prototype (think about counterfeit money or the disguised skunk).
We need to ask, therefore, how you think about categories when you’re
not guided by typicality. The answer, it seems, is that you focus on attributes
that you believe are essential for each category. Your judgments about what’s
essential, however, depend, on your beliefs about that category. Therefore,
you consider parentage when thinking about a category (like skunk or
raccoon) for which you believe biological inheritance is important. You
consider circumstances of printing when you’re concerned with a category
The Difficulties with Categorizing via Resemblance
•
341
(like counterfeit money) that’s shaped by your beliefs about economic systems. And so on.
Is it possible, though, that we’re pushed into these complexities only because
we’ve been discussing oddball categories such as abused citrus fruits and transformed forest animals? The answer is no, because similar complexities emerge
in less exotic cases. The reason is that the prototype and exemplar views both
depend on judgments of resemblance (resemblance either to a prototype or to
some remembered instance), and resemblance, in turn, is itself a complex notion.
How do you decide whether two objects resemble each other? The obvious suggestion is that objects resemble each other if they share properties,
and the more properties shared, the greater the resemblance. Therefore, we
can say there’s some resemblance between an apple and a tennis ball because
they share a shape (round) and a size (about 3 or 4 inches in diameter). The
resemblance is limited, though, because there are many properties that these
objects don’t share (color, “furry” surface, and so on).
It turns out, though, that this idea of “resemblance from shared properties” won’t work. To see why, consider plums and lawn mowers; how much
do these two things resemble each other? Common sense says they don’t
resemble each other at all, but we’ll reach the opposite conclusion if we simply
count “shared properties” (Murphy & Medin, 1985). After all, both weigh
less than a ton, both are found on Earth, both have a detectable odor, both
are used by people, both can be dropped, both cost less than a thousand dollars, both are bigger than a grain of sand, both are unlikely birthday presents
for your infant daughter, both contain carbon molecules, both cast a shadow
on a sunny day. And on and on and on. With a little creativity, you could
probably count thousands of properties shared by these two objects — but
that doesn’t change the basic assessment that there’s not a close resemblance here. (For discussion, see Goldstone & Son, 2012; Goodman, 1972;
Markman & Gentner, 2001; Medin, Goldstone, & Gentner, 1993.)
The solution to this puzzle, though, seems easy: Resemblance does depend
on shared properties, but — more precisely — it depends on whether the
objects share important, essential properties. On this basis, you regard plums
and lawn mowers as different from each other because the features they share
are trivial or inconsequential. But this idea leads to a question: How do you
decide which features to ignore when assessing similarity and which features
to consider? How do you decide, in comparing a plum and a lawn mower,
which features are relevant and which ones aren’t?
These questions bring us back to familiar territory, because your decisions
about which features are important depend on your beliefs about the concept in question. Thus, in judging the resemblance between plums and lawn
mowers, you were unimpressed that they share the feature “cost less than a
thousand dollars.” That’s because you believe cost is irrelevant for these categories. (If a super-deluxe lawn mower cost a million dollars, it would still be
a lawn mower, wouldn’t it?) Likewise, you don’t perceive plums to be similar
to lawn mowers even though both weigh less than a ton, because you know
this attribute, too, is irrelevant for these categories.
342 •
C H A P T E R N I N E Concepts and Generic Knowledge
Overall, then, the idea is that prototype use depends on judgments of
resemblance (i.e., resemblance between a candidate object and a prototype). Judgments of resemblance, in turn, depend on your being able to
focus on the features that are essential, so that you’re not misled by trivial
features. And, finally, decisions about what’s essential (cost or weight or
whatever) vary from category to category, and vary in particular according to your beliefs about that category. Thus, cost isn’t essential for plums
and lawn mowers, but it is a central attribute for other categories (e.g., the
category “luxury item”). Likewise, having a particular weight isn’t
essential for plums or lawn mowers, but it is prominent for other categories. (Does a sumo wrestler resemble a hippopotamus? Here you might be
swayed by weight.)
The bottom line is that you’re influenced by your background beliefs
when considering oddball cases like the mutilated lemon. But you’re also
influenced by your beliefs in ordinary cases, including, we now see, any case
in which you’re relying on a judgment of resemblance.
TEST YOURSELF
7.Give an example in
which something is
definitely a category
member even though
it has little resemblance to the prototype for the category.
8.In judging similarity,
why is it not enough
simply to count all of
the properties that
two objects have in
common?
Concepts as Theories
It seems clear, then, that our theorizing needs to include more than prototypes
and exemplars. Several pieces of evidence point this way, including the fact
that whenever you use a prototype or exemplar, you’re relying on a judgment
of resemblance, and resemblance, we’ve argued, depends on other knowledge —
knowledge about which attributes to pay attention to and which ones to
regard as trivial. But what is this other knowledge?
BLUE GNU
In judging resemblance or in categorizing an object,
you focus on the features that you believe are
important for an object of that type, and you ignore
nonessential features. Imagine that you encounter a
creature and wonder what it is. Perhaps you reason,
“This creature reminds me of the animal I saw in the
zoo yesterday. The sign at the zoo indicated that
the animal was a gnu, so this must be one, too. Of
course, the gnu in the zoo was a different color and
slightly smaller. But I bet that doesn’t matter. Despite
the new blue hue, this is a zoo gnu, too.” Notice that
in drawing this conclusion you’ve decided that color
isn’t a critical feature, so you categorize despite
the contrast on this dimension. But you know that
color does matter for other categories — and so, for
example, you know that something’s off if a jeweler
tries to sell you a green ruby or a red emerald. Thus,
in case after case, the features that you consider
depend on the specific category.
Concepts as Theories
•
343
Explanatory Theories
In the cases we’ve discussed, your understanding of a concept seems to
involve a network of beliefs linking the target concept to other concepts. To
understand what counterfeit is, you need to know what money is, and probably what a government is, and what crime is. To understand what a raccoon
is, you need to understand what parents are, and with that, you need to know
some facts about life cycles, heredity, and the like.
Perhaps, therefore, we need to change our overall approach. We’ve been
trying throughout this chapter to characterize concepts one by one, as
though each concept could be characterized independently of other concepts.
We talked about the prototype for bird, for example, without considering
how this prototype is related to the animal prototype or the egg prototype.
Maybe, though, we need a more holistic approach, one in which we put more
emphasis on the interrelationships among concepts. This would enable us to
include in our accounts the wide network of beliefs in which concepts seem
to be embedded.
To see how this might play out, let’s again consider the concept “raccoon.” Your knowledge about this concept probably includes a raccoon
prototype and some exemplars, and you rely on these representations in
many settings. But your knowledge also includes your belief that raccoons
are biological creatures (and therefore the offspring of adult raccoons) and
your belief that raccoons are wild animals (and therefore usually not pets,
usually living in the woods). These various beliefs may not be sophisticated,
and they may sometimes be inaccurate, but nonetheless they provide you
with a broad cause-and-effect understanding of why raccoons are as they
are. (Various authors have suggested different proposals for how we should
conceptualize this web of beliefs. See, e.g., Bang, Medin, & Atran, 2007;
Keil, 1989, 2003; Lakoff, 1987; Markman & Gentner, 2001; Murphy, 2003;
Rips et al., 2012.)
Guided by these considerations, many authors suggest that each of us has
something we can think of as a “theory” about raccoons — what they are,
how they act, why they are as they are — and likewise a “theory” about most
of the other concepts we hold. The theories are less precise, less elaborate,
than a scientist’s theory, but they serve the same function. They provide a
crucial knowledge base that we rely on in thinking about an object, event,
or category; and they enable us to understand new facts we might encounter
about the object or category.
The Function of Explanatory Theories
We’ve already suggested that implicit “theories” influence how you categorize
things — that is, your decisions about whether a test case is or is not in a particular category. This was crucial in our discussion of the abused lemon, the
transformed raccoon, and the counterfeit bill. Your “theory” for a concept
was also crucial for our discussion of resemblance — guiding your decisions
about which features matter in judging resemblance and which ones do not.
344 •
C H A P T E R N I N E Concepts and Generic Knowledge
As a different example, imagine that you see someone at a party jump
fully clothed into a pool. Odds are good that you would decide this person
belongs in the category “drunk,” but why? Jumping into a pool in this way
surely isn’t part of the definition of being drunk, and it’s unlikely to be part
of the prototype (Medin & Ortony, 1989). But each of us has certain beliefs
about how drunks behave; we have, in essence, a “theory” of drunkenness.
This theory enables us to think through what being drunk will cause someone to do, and on this basis we would decide that, yes, someone who jumped
into the pool fully clothed probably was drunk.
You also draw on a “theory” when thinking about new possibilities for a
category. For example, could an airplane fly if it were made of wood? What if
it were ceramic? How about one made of whipped cream? You immediately
reject this last option, because you know that a plane’s function depends on
its aerodynamic properties, and those depend on the plane’s shape. Whipped
cream wouldn’t hold its shape, so it isn’t a candidate for airplane construction. This is an easy conclusion to draw — but only because your “airplane”
concept contains some ideas about why airplanes are as they are.
Your “theories” also affect how quickly you learn new concepts. Imagine
that you’re given a group of objects and must decide whether each belongs
in Category A or Category B. Category A, you’re told, includes objects that
are metal, have a regular surface, are of medium size, and are easy to grasp.
Category B, in contrast, includes objects that aren’t made of metal, have
irregular surfaces, and are small and hard to grasp. This sorting task would
be difficult — unless we give you another piece of information: namely, that
A WOODEN AIRPLANE?
Could an airplane be made of
wood? Made from ceramic?
Made from whipped cream? You
immediately reject the last possibility, because your “theory”
about airplanes tells you that
planes can fly only because of
their wings’ shape, and whipped
cream wouldn’t maintain this
shape. Planes can, however, be
made of wood — and this one
(the famous Spruce Goose) was.
Concepts as Theories
•
345
Category A includes objects that could serve as substitutes for a hammer.
With this clue, you immediately draw on your other knowledge about hammers, including your understanding of what a hammer is and how it’s used.
This understanding enables you to see why Category A’s features aren’t an
arbitrary hodgepodge; instead, the features form a coherent package. And
once you see this point, learning the experimenter’s task (distinguishing
Category A from Category B) is easy. (See Medin, 1989; Wattenmaker, Dewey,
Murphy, & Medin, 1986. For related findings, see Heit & Bott, 2000; Kaplan
& Murphy, 2000; Rehder & Ross, 2001.)
Inferences Based on Theories
If you meet my pet, Milo, and decide that he’s a dog, then you instantly know
a great deal about Milo — the sorts of things he’s likely to do (bark, beg for
treats, chase cats) and the sorts of things he’s unlikely to do (climb trees, play
chess, hibernate all winter). Likewise, if you learn some new fact about Milo,
you’ll be able to make broad use of that knowledge — applying it to other
creatures of his kind. If, for example, you learn that Milo is vulnerable to circovirus, you’ll probably conclude that other dogs are also vulnerable to this virus.
These examples remind us of one of the reasons categorization is so
important: Categorization enables you to apply your general knowledge (e.g.,
knowledge about dogs) to new cases you encounter (e.g., Milo). Conversely,
categorization enables you to draw broad conclusions from your experience
(so that things you learn about Milo can be applied to other dogs you meet).
All this is possible, though, only because you realize that Milo is a dog; without this simple realization, you wouldn’t be able to use your knowledge in
this way. But how exactly does this use-of-knowledge proceed?
Early research indicated that inferences about categories were guided by
typicality. In one study, participants, told a new fact about robins, were willing to infer that the new fact would also be true for ducks. If they were told
a new fact about ducks, however, they wouldn’t extrapolate to robins (Rips,
1975). Apparently, people were willing to make inferences from the typical
case to the whole category, but not from an atypical case to the category. (For
discussion of why people are more willing to draw conclusions from typical
cases, see Murphy & Ross, 2005.)
However, your inferences are also guided by your broader set of beliefs,
and so, once again, we find a role for the “theory” linked to each concept.
For example, if told that gazelle’s blood contains a certain enzyme, people are
willing to conclude that lion’s blood contains the same enzyme. However, if
told that lion’s blood contains the enzyme, people are less willing to conclude
that gazelle’s blood does too. What’s going on here? People find it easy to
believe the enzyme can be transmitted from gazelles to lions, because they
can easily imagine that lions sometimes eat gazelles; people have a harder
time imagining a mechanism that would transmit the enzyme in the reverse
direction. Likewise, if told that grass contains a certain chemical, people are
willing to believe that cows have the same chemical inside them. This makes
346 •
C H A P T E R N I N E Concepts and Generic Knowledge
WHY IS CATEGORIZATION SO IMPORTANT?
If you decide that Milo is a dog, then you instantly know a great deal
about him (e.g., that he’s likely to bark and chase cats, unlikely to climb
trees or play chess). In this way, categorization enables you to apply
your general knowledge to new cases. And if you learn something
new about Milo (e.g., that he’s at risk for a particular virus), you’re likely
to assume the same is true for other dogs. In this way, categorization
also enables you to draw broad conclusions from specific experiences.
perfect sense if people are thinking of the inference in terms of cause and
effect, relying on their beliefs about how these concepts are related to each
other (Medin, Coley, Storms, & Hayes, 2003; also see Heit, 2000; Heit &
Feeney, 2005; Rehder & Hastie, 2004).
Different Profiles for Different Concepts
This proposal about “theories” and background knowledge has another
implication: People may think about different concepts in different ways. For
example, most people believe that natural kinds (groups of objects that exist
naturally in the world, such as bushes or alligators or stones or mountains) are
as they are because of forces of nature that are consistent across the years. As a
result, the properties of these objects are relatively stable. Thus there are certain
properties that a bush must have in order to survive as a bush; certain properties that a stone must have because of its chemical composition. Things are
different, though, for artifacts (objects made by human beings). If we wished to
make a table with 15 legs rather than 4, or one made of gold, we could do this.
The design of tables is up to us; and the same is true for most artifacts.
This observation leads to the proposal that people will reason differently
about natural kinds and artifacts — because they have different beliefs about
why categories of either sort are as they are. We’ve already seen one result
consistent with this idea: the finding that children agree that toasters could
be turned into coffeepots but not that skunks could be turned into raccoons.
Plainly, the children had different ideas about artifacts (like toasters) than
they had about animate objects (like skunks). Other results confirm this pattern. In general, people tend to assume more stability and more homogeneity
Concepts as Theories
•
347
when reasoning about natural kinds than when reasoning about artifacts
(Atran, 1990; Coley, Medin, & Atran, 1997; Rehder & Hastie, 2004).
The diversity of concepts, as well as the role of beliefs, is also evident in
another context. Many concepts can be characterized in terms of their features (e.g., the features that most dogs have, the features that chairs usually
have, and so on; after Markman & Rein, 2013). Other concepts, though,
involve goal-derived categories, like “diet foods” or “exercise equipment”
(Barsalou, 1983, 1985). Your understanding of concepts like these depends
on your understanding of the goal (e.g., “losing weight”) and some causeand-effect beliefs about how a particular food might help you achieve that
goal. Similar points apply to relational categories (“rivalry,” “hunting”) and
event categories (“visits,” “dates,” “shopping trips”); here, too, you’re influenced by a web of beliefs about how various elements (the predator and the
prey; the shopper and the store) are related to each other.
Concepts and the Brain
The contrasts among different types of concepts are also reflected in neuroscience evidence. For example, fMRI scans show that different brain sites are
activated when people are thinking about living things than when thinking
about nonliving things (e.g., Chao, Weisberg, & Martin, 2002), and different
sites are activated when people are thinking about manufactured objects such
as tools rather than natural objects such as rocks (Gerlach, Law, & Paulson,
2002; Kellenbach et al., 2003).
These results suggest that different types of concepts are represented in different brain areas, and this point is confirmed by observations of people who
have suffered brain damage. In some cases, these people lose the ability to
name certain objects — a pattern termed anomia — or to answer simple questions about these objects (“Does a whale have legs?”). Often, the problem is
specific to certain categories, such that some patients lose the ability to name
living things but not nonliving things; other patients show the reverse pattern. (See Mahon & Caramazza, 2009; Mahon & Hickok, 2016. For broader
discussion, see Peru & Avesani, 2008; Phillips, Noppeney, Humphreys, &
Price, 2002; Rips et al., 2012; Warrington & Shallice, 1984.) Sometimes the
disruption caused by brain damage is even more specific, with some patients
losing the ability to answer questions about fruits and vegetables but still able
to answer questions about other objects, living or nonliving (see Figure 9.4).
Why does the brain separate things in this way? One proposal emphasizes
the idea that different types of information are essential for different concepts.
In this view, the recognition of living things may depend on perceptual properties (especially visual properties) that allow us to identify horses or trees or
other animate objects. In contrast, the recognition of nonliving things may
depend on their functional properties (Warrington & McCarthy, 1983, 1987).
As an interesting complication, though, brain scans also show that sensory
and motor areas in the brain are activated when people are thinking about
certain concepts (Mahon & Caramazza, 2009; Mahon & Hickok, 2016;
348 •
C H A P T E R N I N E Concepts and Generic Knowledge
FIGURE 9.4
DIFFERENT BRAIN SITES SUPPORT DIFFERENT CATEGORIES
Lesion data
Lesion results summarized
Persons
Animals
TP
Tools
A
IT
IT+
Persons: x = 59.8
Persons: x = 75.5
Persons: x = 91.7
Animals: x = 93.3
Animals: x = 80.1
Animals: x = 88.3
Tools: x = 96.0
Tools: x = 84.5
Tools: x = 78.5
B
Brain damage often causes anomia — an inability to name common objects. But the specific loss depends on
where exactly the brain damage has occurred. Panel A summarizes lesion data for patients who had difficulty
naming persons (top), animals (middle), or tools (bottom). The colors indicate the percentage of patients with
damage at each site: red, most patients; purple, few. Panel B offers a different summary of the data: Patients with
damage in the brain’s temporal pole (TP, shown in blue) had difficulty naming persons (only 59.8% correct) but
were easily able to name animals and tools. Patients with damage in the inferotemporal region (IT, shown in red)
had difficulty naming persons and animals but did somewhat better naming tools. Finally, patients with damage in
the lateral occipital region (IT+) had difficulty naming tools but did reasonably well naming animals and persons.
( after damasio , grabowski , tranel , hichwa , & damasio , 1996)
McRae & Jones, 2012). For example, when someone is thinking about the
concept “kick,” we can observe activation in brain areas that (in other circumstances) control the movement of the legs; when someone is thinking about
rainbows, we can detect activation in brain areas ordinarily involved in color
vision. Findings like these suggest that conceptual knowledge is intertwined
with knowledge about what particular objects look like (or sound like or feel
like) and also with knowledge about how one might interact with the object.
Concepts as Theories
•
349
TEST YOURSELF
9.Why is an (informal,
usually unstated)
“theory” needed in
judging the resemblance between
two objects?
10.What’s different
between your (informal, usually unstated)
theory of artifacts and
your theory of natural
kinds?
Some theorists go a step further and argue for a position referred to
as “embodied” or “grounded cognition.” The proposal is that the body’s
sensory and action systems play an essential role in all our cognitive
processes; it’s inevitable, then, that our concepts will include representations of perceptual properties and motor sequences associated with each
concept. (See Barsalou, 2008, 2016; Chrysikou, Csasanto, & ThompsonSchill, 2017; Pulvermüller, 2013. For a glimpse of the debate over this perspective, see Binder, 2016; Bottini, Bucur, & Crepaldi, 2016; Dove, 2016;
Goldinger, Papesh, Barnhart, Hansen, & Hout, 2016; Leshinskaya &
Caramazza, 2016; Reilly, Peele, Garcia, & Crutch, 2016. For a discussion of
how this approach might handle abstract concepts, see Borghi et al., 2017.)
Even if we hold the embodied cognition proposal to the side, the data
here fit well with a theme we’ve been developing throughout this chapter —
namely, that conceptual knowledge has many elements. These include a prototype, exemplars, a theory, and (we now add) representations of perceptual
properties and actions associated with the concept. Let’s also emphasize
that which of these elements you’ll focus on likely depends on your needs at
that moment. In other words, each of your concepts includes many types of
information, but when you’re using your conceptual knowledge, you call to
mind just the subset of information that’s needed for whatever task you’re
engaged in. (See Mahon & Hickok, 2016, especially pp. 949–950; also Yee
& Thompson-Schill, 2016.)
The Knowledge Network
Overall, our theorizing is going to need some complexities, but, within this
complexity, one idea has come up again and again: How you think about
your concepts, how you use your concepts, and what your concepts are, are
all shaped by a web of beliefs and background knowledge. But what does this
“web of beliefs” involve?
Traveling through the Network
to Retrieve Knowledge
In earlier chapters, we explored the idea that information in long-term
memory is represented by means of a network, with associative links connecting nodes to one another. Let’s now carry this proposal one step further.
The associative links don’t just tie together the various bits of knowledge;
they also help represent the knowledge. For example, you know that George
Washington was an American president. This simple idea can be represented
as an associative link between a node representing washington and a node
representing president. In other words, the link itself is a constituent of the
knowledge.
On this view, how do you retrieve knowledge from the network, so that
you can use what you know? Presumably, the retrieval relies on processes
we’ve described in other chapters — with activation spreading from one
350 •
C H A P T E R N I N E Concepts and Generic Knowledge
node to the next. This spread of activation is quick but does take time, and
the farther the activation must travel, the more time needed. This leads to a
prediction — that you’ll need less time to retrieve knowledge involving closely
related ideas, and more time to retrieve knowledge about more distant ideas.
Collins and Quillian (1969) tested this prediction many years ago, using the
sentence verification task described earlier in this chapter. Their participants
were shown sentences such as “A robin is a bird” or “Cats have claws” or
“Cats have hearts.” Mixed together with these obviously true sentences were
various false sentences (e.g., “A cat is a bird”), and in response to each sentence,
participants had to hit a “true” or “false” button as quickly as they could.
Participants presumably perform this task by “traveling” through the network, seeking a connection between nodes. When the participant finds the
connection from, say, the robin node to the birds node, this confirms that
there’s an associative path linking these nodes, which tells the participant
that the sentence about these two concepts is true. This travel should require
little time if the two nodes are directly linked by an association, as robin and
birds probably are (see Figure 9.5). In this case, we’d expect participants
to answer “true” rather quickly. The travel will require more time, however,
if the two nodes are connected only indirectly (e.g., robin and animals),
FIGURE 9.5
HYPOTHETICAL MEMORY STRUCTURE FOR KNOWLEDGE ABOUT ANIMALS
HAVE HEARTS
ANIMALS
BIRDS
CAN FLY
CATS
LAY EGGS
ROBIN
CANARY
EAT FOOD
BREATHE
HAVE SKIN
DOGS
HAVE CLAWS
CHESHIRE
ALLEY CAT
CHASE CATS
BARK
PURR
COLLIE
TERRIER
CAN SING
IS YELLOW
(etc.)
Collins and Quillian proposed that the memory system avoids redundant storage of connections between cats and
have hearts, and between dogs and have hearts, and so on for all the other animals. Instead, have hearts is stored
as a property of all animals. To confirm that cats have hearts, therefore, you must traverse two links: from cats to
animals, and from animals to have hearts.
( after collins & quillian , 1969)
The Knowledge Network
•
351
so that we’d expect slower responses to sentences that require a “two-step”
connection than to sentences that require a single connection.
Collins and Quillian also argued that there’s no point in storing in
memory the fact that cats have hearts and the fact that dogs have hearts
and the fact that squirrels have hearts. Instead, they proposed, it would
be more efficient just to store the fact that these various creatures are animals, and then the separate fact that animals have hearts. As a result, the
property “has a heart” would be associated with the animals node rather
than the nodes for each individual animal, and the same is true for all
the other properties of animals, as shown in the figure. According to this
logic, we should expect relatively slow responses to sentences like “Cats
have hearts,” since, to choose a response, a participant must locate the
linkage from cat to animals and then a second linkage from animals to
have hearts. We would expect a quicker response to “Cats have claws,”
because here there would be a direct connection between cat and the node
representing this property. (Why a direct connection? All cats have claws
but some other animals don’t, so this information couldn’t be entered at
the higher level.)
As Figure 9.6 shows, these predictions are borne out. Responses to
sentences like “A canary is a canary” take approximately 1 second (1,000 ms).
This is presumably the time it takes just to read the sentence and to move
your finger on the response button. Sentences like “A canary can sing”
require an additional step of traversing one link in memory and yield slower
responses. Sentences like “A canary can fly” require the traversing of two
links, from canary to birds and then from birds to can fly, so they are
correspondingly slower.
More recent data, however, add some complications. For example, we
saw earlier in the chapter that verifications are faster if a sentence involves
creatures close to the prototype — so that responses are faster to, say, “A
canary is a bird” than to “An ostrich is a bird.” This difference isn’t reflected
in Figure 9.6, nor is it explained by the layout in Figure 9.5. Clearly, then,
the Collins and Quillian view is incomplete.
In addition, the principle of “nonredundancy” proposed by Collins and
Quillian doesn’t always hold. For example, the property of “having feathers”
should, on their view, be associated with the birds node rather than
(redundantly) with the robin node, the pigeon node, and so on. This fits
with the fact that responses are relatively slow to sentences like “Sparrows
have feathers.” However, it turns out that participants respond rather quickly
to a sentence like “Peacocks have feathers.” This is because in observing peacocks, you often think about their prominent tail feathers (Conrad, 1972).
Therefore, even though it is informationally redundant, a strong association
between peacock and have feathers is likely to be established.
Even with these complications, we can often predict the speed of knowledge access by counting the number of nodes participants must traverse in
answering a question. This observation powerfully confirms the claim that
associative links play a pivotal role in knowledge representation.
352 •
C H A P T E R N I N E Concepts and Generic Knowledge
FIGURE 9.6
T IME NEEDED TO CONFIRM VARIOUS
SEMANTIC FACTS
A canary
has skin.
1,500
Mean response time (ms)
1,400
A canary
can sing.
1,300
A canary
is a bird.
1,200
1,100
A canary
can fly.
A canary is
an animal.
A canary is
a canary.
1,000
Property
900
Category
1
0
2
Levels to be traversed
In a sentence verification task, participants’ responses were fastest when the
test required them to traverse zero links in memory (“A canary is a canary”),
slower when the necessary ideas were separated by one link, and slower still
if the ideas were separated by two links. Responses were also slower if participants had to take the additional step of traversing the link from a category
label (“bird”) to the node representing a property of the category (can fly).
( after collins & quillian , 1969)
Propositional Networks
To represent the full fabric of your knowledge, however, we need more than
simple associations. After all, we need somehow to represent the contrast
between “Sam has a dog” and “Sam is a dog.” If all we had is an association
between sam and dog, we wouldn’t be able to tell these two ideas apart.
One widely endorsed proposal solves this problem with a focus on propositions, defined as the smallest units of knowledge that can be either true or
false (Anderson, 1976, 1980, 1993; Anderson & Bower, 1973). For example,
“Children love candy” is a proposition, but “Children” is not; “Susan likes
blue cars” is a proposition, but “blue cars” is not. Propositions are easily represented as sentences, but this is just a convenience. They can also be represented
in various nonlinguistic forms, including a structure of nodes and linkages, and
that’s exactly what Anderson’s model does.
The Knowledge Network
•
353
Figure 9.7 provides an example. Here, each ellipse identifies a single proposition. Associations connect an ellipse to ideas that are the proposition’s constituents, and the associations are labeled to specify the constituent’s role within that
proposition. This enables us to distinguish, say, the proposition “Dogs chase
cats” (shown in the figure) from the proposition “Cats chase dogs” (not shown).
This model shares many claims with the network theorizing we discussed
in earlier chapters. Nodes are connected by associative links. Some of these
links are stronger than others. The strength of a link depends on how frequently and recently it has been used. Once a node is activated, the process
of spreading activation causes nearby nodes to become activated as well. The
model is distinctive, however, in its attempt to represent knowledge in terms
of propositions, and the promise of this approach has attracted the support
of many researchers. (For recent discussion, see Salvucci, 2017; for some
alternative models, see Flusberg & McClelland, 2017; Kieras, 2017. For
more on how this network can store information, see Figure 9.8.)
Distributed Processing
In the model just described, individual ideas are represented with local representations. Each node represents one idea so that when that node is activated,
you’re thinking about that idea, and when you’re thinking about that idea,
that node is activated. Connectionist networks, in contrast, take a different
approach. They rely on distributed representations, in which each idea is represented, not by a certain set of nodes, but instead by a pattern of activation
across the network. To take a simple case, the concept “birthday” might be
represented by a pattern in which nodes b, f, h, n, p, and r are firing, whereas
the concept “computer” might be represented by a pattern in which nodes
c, g, h, m, o, and s are firing. Note that node h is part of both of these patterns and probably part of the pattern for other concepts as well. Therefore,
we can’t attach any meaning or interpretation to this node by itself; we can
only learn what’s being represented by looking at many nodes simultaneously
to find out what pattern of activation exists across the entire network. (For
more on local and distributed representations, see Chapter 4; also see the
related discussion of neural coding in Chapter 2.)
This reliance on distributed representation has important consequences
for how a connectionist network functions. Imagine being asked what sort
of computer you use. For you to respond, the idea “computer” needs to
trigger the idea “MacBook” (or “Toshiba” or whatever it is you have). In a
distributed network, this means that the many nodes representing the concept “computer” have to manage collectively to activate the many nodes
representing “MacBook.” To continue our simple illustration, node c has to
trigger node l at the same time that node g triggers node a, and so on, leading ultimately to the activation of the l-a-f-j-t-r combination that, let’s say,
represents “MacBook.” In short, a network using distributed representations
must use processes that are similarly distributed, so that one widespread
activation pattern can evoke a different (but equally widespread) pattern.
354 •
C H A P T E R N I N E Concepts and Generic Knowledge
FIGURE 9.7
N
ETWORK REPRESENTATIONS OF SOME
OF YOUR KNOWLEDGE ABOUT DOGS
CHEW
Relation
DOG
Agent
BONE
Agent
Agent
CHASE
Object
Subject
Relation PART
OF
Relation
Object Object
Relation
Object
CAT
EAT
MEAT
Your understanding of dogs — what they are, what they’re likely to do — is
represented by an interconnected network of propositions, with each proposition being indicated by an ellipse. Labels on the arrows indicate each node’s
role within the proposition.
( after anderson , 1980)
FIGURE 9.8
EPRESENTING EPISODES WITHIN
R
A PROPOSITIONAL NETWORK
LAST SPRING
Time
JACOB
Agent
Object
Relation
FEEDS
PIGEONS
IN
Relation
Location
TRAFALGAR
SQUARE
In order to represent episodes, the propositional network includes time and
location nodes. This fragment of a network represents two propositions: the
proposition that Jacob fed pigeons last spring, and the proposition that the
pigeons are in Trafalgar Square. Notice that no time node is associated with
the proposition about pigeons being in Trafalgar Square. Therefore, what’s
represented is that the feeding of the pigeons took place last spring but that
the pigeons are always in the square.
The Knowledge Network
•
355
COGNITION
outside the lab
Stereotypes
You have concepts of things (“chair,” “book,”
efficient means of organizing large quantities of
“kitchen”),
(“running,”
information (and so both serve the same function
“hiding,” “dancing”), and concepts for animals
as schemata, which we described in Chapter 8). But,
(“cat,” “cow,” “dragon”). But you also have con-
of course, a reliance on stereotypes can lead to a
cepts that apply to people. You understand what
list of toxic problems — racism, sexism, homophobia,
a “nurse” is, and a “toddler,” and a “nerd” or a
prejudice against anyone wearing a hijab, and more.
“jock.” You also have concepts for various reli-
Many factors fuel these ugly tendencies,
gious, racial, and ethnic groups (“Jew,” “Muslim,”
including the fact that people often act as if all
“African American,” “Italian”), various political
members of the stereotyped group are alike. They
groups (“radical,” “ultra-conservative”), and many
assume, for example, that a tall African American
others as well.
individual is probably a talented basketball player,
concepts
for
actions
In many ways, concepts representing your ideas
and that a Semitic-looking young man wearing a
about groups of people have the same profile as any
headscarf is probably a terrorist. These assump-
other concepts: It’s difficult to find a rigid definition
tions are, of course, indefensible because humans
for most of these groups, because we can usually
in any group differ from one another, and there’s
find individuals who are in the group even though
no justification for jumping to conclusions about
they don’t quite fit the definition. You also have a
someone just because you’ve decided he or she is
cluster of interwoven beliefs (a “theory”) about these
a member of a particular group.
groups — beliefs that link your ideas about the group
This kind of assumption, though, is widespread
to many other ideas. You also have a prototype in
enough so that social psychologists give it a name:
mind for the group, but here we typically use a dif-
the outgroup homogeneity effect. This term refers
ferent term: You have a stereotype for the group.
to the fact that most people are convinced that their
How are stereotypes different from prototypes?
“ingroup” (the group they belong to) is remarkably
Prototypes are a summary of your experience —
varied, while “outgroups” (groups they don’t belong
and so your prototype for “dog” can be thought
to) are quite homogeneous. In other words, no mat-
of as an average of all the dogs you’ve seen.
ter who you count as “they” and who you count as
Stereotypes, in contrast, are often acquired
“we,” you’re likely to agree that “they all think and
through social channels — with friends or family,
act alike; we, however, are wonderfully diverse.”
or perhaps public figures, shaping your ideas
In combating prejudice, then, it’s useful to real-
about what “lawyers” are like, or “Canadians,” or
ize that this assumption of homogeneity isn’t just
“Italians.” In addition, stereotypes often include an
wrong; it can also have ugly consequences. There
emotional or evaluative dimension, with the result
may be intellectual efficiency in thinking about wom-
that there are groups you’re afraid of, groups you
en, or the elderly, or politicians as if these groups
respect, groups you sneer at.
were uniform, but in doing so you fail to respect
Let’s acknowledge, though, that stereotypes can
the differences from one person to the next — and
serve the same cognitive function as prototypes.
may end up with beliefs, feelings, or actions that are
In both cases, these representations provide an
impossible to justify and often deeply harmful.
356 •
C H A P T E R N I N E Concepts and Generic Knowledge
In addition, the steps bringing this about must all occur simultaneously — in
parallel — with each other, so that one entire representation can smoothly
trigger the next. This is why connectionist models are said to involve parallel
distributed processing (PDP).
Many theorists argue that models of this sort make biological sense. We
know that the brain relies on parallel processing, with ongoing activity in
many regions simultaneously. We also know that the brain uses a “divide and
conquer” strategy, with complex tasks being broken down into small components, and with separate brain areas working on each component.
In addition, PDP models are remarkably powerful, and computers relying
on this sort of processing are often able to conquer problems that seemed
insoluble with other approaches. As a related point, PDP models have an
excellent capacity for detecting patterns in the input they receive, despite a
range of variations in how the pattern is implemented. The models can therefore recognize a variety of different sentences as all having the same structure,
and a variety of game positions as all inviting the same next move. As a result,
these models are impressively able to generalize what they have “learned” to
new, never-seen-before variations on the pattern. (For a broad view of what
connectionism can accomplish, see Flusberg & McClelland, 2017.)
Learning as the Setting of Connection Weights
How do PDP models manage to detect patterns? How do these models “learn”?
Recall that in any associative network, knowledge is represented by the associations themselves. To return to an earlier example, the knowledge that “George
Washington was president” is represented via a link between the nodes representing “Washington” and those representing “president.” When we first introduced
this example, we phrased it in terms of local representations, with individual
nodes having specific referents. The idea, however, is the same in a distributed
system. What it means to know this fact about Washington is to have a pattern
of connections among the many nodes that together represent “Washington”
and the many nodes that together represent “president.” Once these connections
are in place, activation of either pattern will lead to the activation of the other.
Notice, then, that knowledge refers to a potential rather than to a state. If
you know that Washington was a president, then the connections are in place
so that if the “Washington” pattern of activations occurs, this will lead to the
“president” pattern of activations. And this state of readiness will remain even
if you happen not to be thinking about Washington right now. In this way,
“knowing” something, in network terms, corresponds to how the activation
will flow if there is activation on the scene. This is different from “thinking
about” something, which corresponds to which nodes are active at a particular moment, with no comment about where that activation will spread next.
According to this view, “learning” involves adjustments of the connections
among nodes, so that after learning, activation will flow in a way that can represent the newly gained knowledge. Technically, we would say that learning
involves the adjustment of connection weights — the strength of the individual
The Knowledge Network
•
357
TEST YOURSELF
11.What does it mean
to say that knowledge can be represented via network
connections?
12.What is a propositional network?
13.Why do distributed
representations
require distributed
processing?
connections among nodes. Moreover, in this type of model, learning requires
the adjustment of many connection weights. We need to adjust the connections, for example, so that the thousands of nodes representing “Washington”
manage, together, to activate the thousands of nodes representing “president.”
In this way, learning, just like everything else in the connectionist scheme, is a
distributed process involving thousands of changes across the network.
Concepts: Putting the Pieces Together
We have now covered a lot of ground — discussing both individual concepts
and also how these concepts might be woven together, via the network, to
form larger patterns of knowledge. We’ve also talked about how the network
itself might be set up — with knowledge perhaps represented by propositions,
or perhaps via a connectionist network. But where does all of this leave us?
You might think there’s nothing glorious or complicated about knowing what
a dog is, or a lemon, or a fish. Your use of these concepts is effortless, and so is your
use of thousands of other concepts. No one over the age of 4 takes special pride
in knowing what an odd number is, nor do people find it challenging to make the
elementary sorts of judgments we’ve considered throughout this chapter.
As we’ve seen, though, human conceptual knowledge is impressively complex. At the very least, this knowledge contains several parts. We’ve suggested
that people have a prototype for most of their concepts as well as a set of
remembered exemplars, and use them for a range of judgments about the
relevant category. People also seem to have a set of beliefs about each concept they hold, and these beliefs reflect the person’s understanding of causeand-effect relationships — for example, why drunks act as they do, or how
enzymes found in gazelles might be transmitted to lions. These beliefs are
woven into the broader network that manages to store all the information in
your memory, and that network influences how you categorize items and also
how you reason about the objects in your world.
Apparently, then, even our simplest concepts require a multifaceted representation in our minds, and at least part of this representation (the “theory”)
seems reasonably sophisticated. It is all this richness, presumably, that makes
human conceptual knowledge extremely powerful and flexible — and so easy
to use in a remarkable range of circumstances.
COGNITIVE PSYCHOLOGY AND EDUCATION
learning new concepts
In your studies, you encounter many new terms. For example, in this book (and
in many others) you’ll find boldfaced terms introducing new concepts, and
often the book provides a helpful definition, perhaps in a glossary (as this book
358 •
C H A P T E R N I N E Concepts and Generic Knowledge
does). As the chapter argues, though, this mode of presentation doesn’t line up
all that well with the structure of human knowledge. The reason is that you
don’t have (or need) a definition for most of the concepts in your repertoire; in
fact, for many concepts, a definition may not even exist. And even when you
do know a definition, your use of the concept often relies on other information —
including a prototype for that term as well as a set of exemplars.
In addition, your use of conceptual information depends on a broader
fabric of knowledge, linking each concept to other things you know. This
broader knowledge encompasses what we’ve called your “theory” about
that concept — a theory that (among other things) explains why the concept’s
attributes are as they are. You use this theory in many ways; for example,
we’ve argued that whenever you rely on a prototype, you’re drawing conclusions based on the resemblance between the prototype and the new
case you’re thinking about, and that resemblance depends on your theory.
Specifically, it’s your theory that tells you which attributes to pay attention to in judging the resemblance, and which ones to ignore. (So if you’re
thinking about computers, for example, your “theory” about computers
tells you that the color of the machine’s case is irrelevant. In contrast, if
you’re identifying types of birds, your knowledge tells you that color is an
important attribute.)
What does all of this imply for the learning of new concepts? First, let’s
be clear that in some technical domains, concepts do have firm definitions.
(For example, in a statistics class, you learn the definition for the mean of a
set of numbers, and that term is precisely defined.) More broadly, though,
you should bear in mind that definitions tell you what’s generally true of a
concept, but rarely name attributes that are always in place. It’s also important not to be fooled into thinking that knowing a definition is the same as
understanding the concept. In fact, if you only know the definition, you may
end up using the concept foolishly. (And so you might misidentify a hairless
Chihuahua: “That couldn’t be a dog — it doesn’t have fur.”)
What other information do you need, in addition to the definition? At the
least, you should seek out examples of the new concept, because you’ll often
be able to draw analogies based on these examples. You also want to think
about what these examples have in common; that will help you develop a
prototype for the category.
In addition, many students (and many teachers) believe that when learning a new concept, it’s best to view example after example, so that you
really master the concept. Then, you can view example after example of the
next concept, so that you’ll learn that one too. But what if you’re trying to
learn about related concepts or categories? What if, for example, you’re an
art student trying to learn what distinguishes Picasso’s artwork from the
work of his contemporaries, or if you’re a medical student learning how
to distinguish the symptom patterns for various diseases? In these settings,
it’s best to hop back and forth with the examples — so that you examine
a couple of instances of this concept, then a couple of instances of that
one, then back to the first, and so on. This interweaving may slow
Cognitive Psychology and Education
•
359
LEARNING NEW
CONCEPTS
When learning to distinguish
two categories, it’s best to
hop back and forth between
the categories. To learn to
distinguish Monet’s art from
van Gogh’s, therefore, you
might view a painting by
Monet, then one by van Gogh,
then another by Monet, and
so on. This sequence will
lead to better learning than a
sequence of first viewing a
large block of Monet’s paintings and then a large block
of van Gogh’s. (The painting
on top is Monet’s The Artist’s
Garden in Argenteuil; the
painting on the bottom is
van Gogh’s Farmhouse in
Provence.)
360 •
down learning initially, but it will help you in the long run (leaving
you with a sharper and longer-lasting understanding) because you’ll learn
both the attributes that are shared within a category and also the attributes
that distinguish one category from another.
In viewing the examples, though, you also want to think about what
makes them count as examples — what is it about them that puts them into
the category? How are the examples different, and why are they all in the
same category despite these differences? Why are other candidates, apparently similar to these examples, not in the category? Are some of the qualities
of the examples predictable from other qualities? What caused these qualities to be as they are? These questions will help you to start building the
network of beliefs that provide your theory about this concept. These beliefs
will help you to understand and use the concept. But, as the chapter discusses,
these beliefs are also part of the concept — providing the knowledge base that
specifies, in your thoughts, what the concept is all about.
These various points put an extra burden on you and your teachers. It
would be easier if the teacher could simply provide a crisp definition for you
to memorize, and then you could go ahead and commit that definition to
memory. But that’s not what it means to learn a concept. Strict attention just
to a definition will leave you with a conceptual representation that’s not very
useful, and certainly far less rich than you want.
For more on this topic . . .
Brown, P. C., Roediger, H. L., & McDaniel, M. A. (2014). Make it stick: The science
of successful learning. New York, NY: Belknap Press.
Dorek, K., Brooks, L., Weaver, B., & Norman, G. (2012). Influence of familiar
features on diagnosis: Instantiated features in an applied setting. Journal of
Experimental Psychology: Applied, 18, 109–125.
Kim, N. S., & Ahn, W.-K. (2002). Clinical psychologists’ theory-based representations of mental disorders predict their diagnostic reasoning and memory.
Journal of Experimental Psychology: General, 131, 451–476.
C H A P T E R N I N E Concepts and Generic Knowledge
chapter review
SUMMARY
• People cannot provide definitions for most of
• Sometimes categorization doesn’t depend at
the concepts they use; this suggests that knowing a
concept and being able to use it competently do not
require knowing a definition. However, when trying to define a term, people mention properties that
are in fact closely associated with the concept. One
proposal, therefore, is that your knowledge specifies
what is typical for each concept, rather than naming properties that are truly definitive for the concept. Concepts based on typicality will have a family
resemblance structure, with different category members sharing features but with no features being
shared by the entire group.
all on whether the test case resembles a prototype
or a category exemplar. This is evident with some
abstract categories (“even number”) and some
weird cases (a mutilated lemon), but it’s also evident
with more mundane categories (“raccoon”). In
these examples, categorization seems to depend on
knowledge about a category’s essential properties.
• Concepts may be represented in the mind via
prototypes, with each prototype representing what
is most typical for that category. This implies that
categories will have graded membership, and many
research results are consistent with this prediction.
The results converge in identifying some category
members as “better” members of the category. This
is reflected in sentence verification tasks, production
tasks, explicit judgments of typicality, and so on.
• In addition, basic-level categories seem to be the
• Knowledge about essential properties is not just a
supplement to categorization via resemblance. Instead,
knowledge about essential properties may be a prerequisite for judgments of resemblance. With this knowledge, you’re able to assess resemblance with regard to
just those properties that truly matter for the category
and not be misled by irrelevant or accidental properties.
• The properties that are essential for a category
vary from one category to the next. The identification of these properties seems to depend on beliefs
held about the category, including causal beliefs that
specify why the category features are as they are.
These beliefs can be thought of as forming implicit
theories, and they describe the category not in isolation but in relation to various other concepts.
ones we learn earliest and use most often. Basiclevel categories (e.g., “chair”) are more homogeneous than their broader, superordinate categories
(“furniture”) and much broader than their subordinate categories (“armchair”). They are also usually
represented by a single word.
• Researchers have proposed that knowledge is
• Typicality results can also be explained with a
• To store all of knowledge, the network may need
model that relies on specific category exemplars,
and with category judgments being made by the
drawing of analogies to these remembered exemplars. The exemplar model can explain your ability
to view categories from a new perspective. Even so,
prototypes provide an efficient summary of what is
typical for the category. Perhaps it’s not surprising,
therefore, that your conceptual knowledge includes
exemplars and prototypes.
stored within the same memory network that we’ve
discussed in earlier chapters. Searching through this
network seems to resemble travel in the sense that
greater travel distances (more connections to be
traversed) require more time.
more than simple associations. One proposal is that the
network stores propositions, with different nodes each
playing the appropriate role within the proposition.
• A different proposal is that knowledge is contained in memory via distributed representations.
These representations require distributed processes,
including the processes that adjust connection
weights to allow the creation of new knowledge.
361
KEY TERMS
family resemblance (p. 329)
prototype theory (p. 329)
graded membership (p. 331)
sentence verification task (p. 331)
production task (p. 331)
rating task (p. 331)
basic-level categorization (p. 333)
exemplar-based reasoning (p. 335)
typicality (p. 337)
anomia (p. 348)
propositions (p. 353)
local representations (p. 354)
connectionist networks (p. 354)
distributed representations (p. 354)
parallel distributed processing (PDP) (p. 357)
connection weights (p. 357)
TEST YOURSELF AGAIN
1.Consider the word “chair.” Name some
attributes that might plausibly be included
in a definition for this word. But, then, can
you describe objects that you would count
as chairs even though they don’t have one or
more of these attributes?
2.What does it mean to say that there is a family
resemblance among the various animals that
we call “dogs”?
3.Why is graded membership a consequence
of representing the category in terms of a
prototype?
4.What tasks show us that concept judgments
often rely on prototypes and typicality?
5.Give an example of a basic-level category, and
then name some of the subcategories within this
basic-level grouping.
6.What is similar in the processes of categorizing
via a prototype and the processes of categorizing via an exemplar? What is different between
these two types of processes?
362
7.Give an example in which something is
definitely a category member even though it
has little resemblance to the prototype for the
category.
8.In judging similarity, why is it not enough
simply to count all of the properties that two
objects have in common?
9.Why is an (informal, usually unstated)
“theory” needed in judging the resemblance
between two objects?
10.What’s different between your (informal,
usually unstated) theory of artifacts and your
theory of natural kinds?
11.What does it mean to say that knowledge can
be represented via network connections?
12. What is a propositional network?
13.Why do distributed representations require
distributed processing?
THINK ABOUT IT
1.You easily understand the following sentence:
“At most colleges and universities, a large
number of students receive financial aid.”
But how do you manage to understand the
sentence? How is the concept of “financial
aid” represented in your mind? Do you have
a prototype (perhaps for “student on financial
aid”)? Do you have some number of exemplars? A theory? Can you specify what the
theory involves? What other concepts do you
need to understand in order to understand
“financial aid”?
E eBOOK DEMONSTRATIONS & ESSAYS
Go to http://digital.wwnorton.com/cognition7 for the online demonstrations and essays
relevant to this chapter. These can be found in the ebook.
Online Demonstrations
• Demonstration 9.1: The Search for Definitions
• Demonstration 9.2: Assessing Typicality
• Demonstration 9.3: Basic-Level Categories
COGNITION LABS
Go to http://digital.wwnorton.com/cognition7 to experience interactive online psychology
labs relevant to this chapter.
363
10
chapter
Language
what if…
On January 8, 2011, Congresswoman Gabby Giffords
was meeting with citizens outside a grocery store near
Tucson, Arizona. A man ran up to the crowd and began shooting. Six
people were killed; Giffords was among the others who were wounded.
A bullet had passed through her head, traveling the length of her brain’s
left side and causing extensive damage.
As a result of her brain injury, Giffords has suffered from many
profound difficulties, including a massive disruption of her language
capacity, and, among its many other implications, her case brought public attention to the disorder termed “aphasia” — a loss of the ability to
produce and understand ordinary language.
In the years since the shooting, though, Giffords has shown a wonderful degree of recovery. Just five months after the injury, an aide
announced that her ability to comprehend language had returned to a
level that was “close to normal, if not normal.” Her progress has been
slower for language production. Consider an interview she gave in early
2014. Giffords had, on the third anniversary of her shooting, decided to
celebrate life by skydiving. In a subsequent TV interview, she described
the experience: “Oh, wonderful sky. Gorgeous mountain. Blue skies. I like
a lot. A lot of fun. Peaceful, so peaceful.”
Giffords’s recovery is remarkable, but — sadly — not typical. The outcome for patients with aphasia is highly variable, and many recover
far less of their language ability than Giffords has. Her case is typical,
though, in other ways. Different brain areas control the comprehension
and the production of speech, so it’s common for one of these capacities
to be spared while the other is damaged, and the prospects for recovery
are generally better for language comprehension than for production.
And like Giffords, many patients with aphasia retain the ability to sing
even if they’ve lost the ability to speak — a clear indication that these
seemingly similar activities are controlled by different processes.
Giffords also shares with other patients the profound frustration of
aphasia. This condition is, after all, a disorder of language, not a disorder of thought. As a result, patients with aphasia can think normally but
complain (often with great difficulty) that they feel “trapped” in their
own heads, unable to express what they’re thinking. They are sometimes
forced to grunt and point in hopes of conveying their meaning; in other
cases, their speech is so slurred that others cannot understand them,
365
preview of chapter themes
•
anguage can be understood as having a hierarchical
L
structure — with units at each level being assembled to
form the larger units at the next level.
•
t each level in the hierarchy, we can combine and recomA
bine units, but the combinations seem to be governed by
various types of rules. The rules provide an explanation of
why some combinations of elements are rare and others
seem prohibited outright. Within the boundaries created by
these rules, though, language is generative, allowing any user
of the language to create a virtually unlimited number of new
forms (new sound combinations, new words, new phrases).
•
different set of principles describes how, moment by
A
moment, people interpret the sentences they encounter;
in this process, people are guided by many factors, including syntax, semantics, and contextual information.
•
In interpreting sentences, people seem to use a “compile
as you go” strategy, trying to figure out the role of each
word the moment it arrives. This approach is efficient but
can lead to error.
•
ur extraordinary skill in using language is made possible
O
in part by the fact that large portions of the brain are specialized for language use, making it clear that we are, in a
literal sense, a “linguistic species.”
•
inally, language surely influences our thoughts, but in an
F
indirect fashion: Language is one of many ways to draw
our attention to this or that aspect of the environment.
This shapes our experience, which in turn shapes our
cognition.
so they are caught in a situation of trying again and again to express
themselves — but often without success.
To understand the extent of this frustration, bear in mind that we use
language (whether it’s the spoken language most of us use or the sign
language of the deaf) to convey our ideas to one another, and our wishes
and our needs. Without language, cooperative endeavors would be a
thousand times more difficult — if possible at all. Without language, the
acquisition of knowledge would be enormously impaired. Plainly, then,
language capacity is crucial for us all, and in this chapter we’ll consider
the nature of this extraordinary and uniquely human skill.
The Organization of Language
Language use involves a special type of translation. I might, for example,
want to tell you about a happy event in my life, and so I need to translate my
ideas about the event into sounds that I can utter. You, in turn, detect those
sounds and need to convert them into some sort of comprehension. How
does this translation — from ideas to sounds, and then back to ideas — take
place?
The answer lies in the fact that language relies on well-defined patterns —
patterns in how individual words are used, patterns in how words are put
together into phrases. I follow those patterns when I express my ideas, and
the same patterns guide you in figuring out what I just said. In essence, then,
we’re both using the same “codebook,” with the result that (most of the time)
you can understand my messages, and I yours.
366 •
C H A P T E R T E N Language
But where does this “codebook” come from? And what’s in the codebook?
More concretely, what are the patterns of English (or whatever language
you speak) that — apparently — we all know and use? As a first step toward
tackling these issues, let’s note that language has a well-defined structure, as
depicted in Figure 10.1. At the highest level of the structure (not shown in
the figure) are the ideas intended by the speaker, or the ideas that the listener
derives from the input. These ideas are typically expressed in sentences —
coherent sequences of words that express the speaker’s intended meaning.
Sentences, in turn, are composed of phrases, which are composed of words.
Words are composed of morphemes, the smallest language units that carry
meaning. Some morphemes, like “umpire” or “talk,” are units that can stand
alone, and they usually refer to particular objects, ideas, or actions. Other morphemes get “bound” onto these “free” morphemes and add information crucial
for interpretation. Examples of bound morphemes in Figure 10.1 are the pasttense morpheme “ed” and the plural morpheme “s.” Then, finally, in spoken
language, morphemes are conveyed by sounds called phonemes, defined as the
smallest unit of sound that serves to distinguish words in a language.
Language is also organized in another way: Within each of these levels,
people can combine and recombine the units to produce novel utterances —
assembling phonemes into brand-new morphemes or assembling words into
THE HIERARCHY OF LINGUISTIC UNITS
The umpires talked to the players
PHRASE
The umpires
WORD
The
MORPHEME
The
PHONEME
i
talked to the players
umpires
umpire
v
mpayr
talked
to
the
the
play
er
s
e
SENTENCE
pley
r
z
s
talk
ed
to
z
t ck
t
tuw
players
e
FIGURE 10.1
It is useful to think of language as having a hierarchical structure. At the top of the hierarchy, there are sentences.
These are composed of phrases, which are themselves composed of words. The words are composed of
morphemes, and when the morphemes are pronounced, the units of sound are called “phonemes.” In describing phonemes, the symbols correspond to the actual sounds produced, independent of how these sounds are
expressed in ordinary writing.
The Organization of Language
•
367
TEST YOURSELF
1.What are morphemes?
What are phonemes?
brand-new phrases. Crucially, though, not all combinations are possible — so
that a new breakfast cereal, for example, might be called “Klof” but would
probably seem strange to English speakers if it were called “Ngof.” Likewise,
someone might utter the novel sentence “I admired the lurking octopi” but
almost certainly wouldn’t say, “Octopi admired the I lurking.” What lies behind
these points? Why are some sequences acceptable — even if strange — while
others seem awkward or even unacceptable? The answers to these questions
are crucial for any understanding of what language is.
Phonology
Let’s use the hierarchy in Figure 10.1 as a way to organize our examination
of language. We’ll start at the bottom of the hierarchy — with the sounds of
speech.
The Production of Speech
In ordinary breathing, air flows quietly out of the lungs and up through the
nose and mouth (see Figure 10.2). There will usually be some sort of sound,
though, if this airflow is interrupted or altered, and this fact is crucial for
vocal communication.
For example, within the larynx there are two flaps of muscular tissue
called the “vocal folds.” (These structures are also called the “vocal cords,”
although they’re not cords at all.) These folds can be rapidly opened and
closed, producing a buzzing sort of vibration known as voicing. You can feel
this vibration by putting your palm on your throat while you produce a [z]
sound. You’ll feel no vibration, though, if you hiss like a snake, producing a
sustained [s] sound. Try it! The [z] sound is voiced; the [s] is not.
You can also produce sound by narrowing the air passageway within the
mouth itself. For example, hiss like a snake again and pay attention to your
tongue’s position. To produce this sound, you placed your tongue’s tip near
the roof of your mouth, just behind your teeth; the [s] sound is the sound of
the air rushing through the narrow gap you created.
If the gap is somewhere else, a different sound results. For example, to
produce the [sh] sound (as in “shoot” or “shine”), the tongue is positioned
so that it creates a narrow gap a bit farther back in the mouth; air rushing
through this gap causes the desired sound. Alternatively, the narrow gap can
be more toward the front. Pronounce an [f] sound; in this case, the sound is
produced by air rushing between your bottom lip and your top teeth.
These various aspects of speech production provide a basis for categorizing speech sounds. We can distinguish sounds, first, according to how
the airflow is restricted; this is referred to as manner of production. Thus,
air is allowed to move through the nose for some speech sounds but not
others. Similarly, for some speech sounds, the flow of air is fully stopped for a
moment (e.g., [p], [b], and [t]). For other sounds, the air passage is restricted,
but air continues to flow (e.g., [f], [z], and [r]).
368 •
C H A P T E R T E N Language
FIGURE 10.2
THE HUMAN VOCAL TRACT
Dental
consonant
region
Palate
Soft palate
Nasal cavity
Oral cavity
Tongue
Lips
Vocal folds
(in the larynx)
Speech is produced by airflow from the lungs that passes through the larynx
and from there through the oral and nasal cavities. Different vowels are created by movements of the lips and tongue that change the size and shape
of the oral cavity. Consonants are produced by movements that temporarily
obstruct the airflow through the vocal tract.
Second, we can distinguish between sounds that are voiced — produced
with the vocal folds vibrating — and those that are not. The sounds of [v],
[z], and [n] (to name a few) are voiced; [f], [s], [t], and [k] are unvoiced. (You
can confirm this by running the hand-on-throat test while producing each of
these sounds.) Finally, sounds can be categorized according to where the airflow is restricted; this is referred to as place of articulation. For example, you
close your lips to produce “bilabial” sounds like [p] and [b]; you place your
top teeth close to your bottom lip to produce “labiodental” sounds like [f]
and [v]; and you place your tongue just behind your upper teeth to produce
“alveolar” sounds like [t] and [d].
This categorization scheme enables us to describe any speech sound in terms
of a few simple features. For example, what are the features of a [p] sound?
Phonology
•
369
First, we specify the manner of production: This sound is produced with air
moving through the mouth (not the nose) and with a full interruption to the
flow of air. Second, voicing: The [p] sound happens to be unvoiced. Third, place
of articulation: The [p] sound is bilabial. These features are all we need to identify the [p], and if any of these features changes, so does the sound’s identity.
In English, these features of sound production are combined and recombined to produce 40 or so different phonemes. Other languages use as few as
a dozen phonemes; still others use many more. (For example, there are 141
different phonemes in the language of Khoisan, spoken by the Bushmen of
Africa; Halle, 1990.) In all cases, though, the phonemes are created by simple
combinations of the features just described.
The Complexity of Speech Perception
This description of speech sounds invites a simple proposal about speech
perception. We’ve just said that each speech sound can be defined in terms of
a small number of features. Perhaps, then, all a perceiver needs to do is detect
these features, and with this done, the speech sounds are identified.
It turns out, though, that speech perception is more complicated. Consider
Figure 10.3, which shows the moment-by-moment sound amplitudes produced by a speaker uttering a brief greeting. It’s these amplitudes, in the form
of air-pressure changes, that reach the ear, and so, in an important sense, the
figure shows the pattern of input with which “real” speech perception begins.
Notice that within this stream of speech there are no markers to indicate
where one phoneme ends and the next begins. Likewise, there are, for the
most part, no gaps to indicate the boundaries between successive syllables
FIGURE 10.3
My
THE ACTUAL PATTERN OF SPEECH
name
is
Dan
Reis
−
berg
Shown here are the moment-by-moment sound amplitudes produced by the author uttering a greeting. Notice
that there is no gap between the sounds carrying the word “my” and the sounds carrying “name.” Nor is there
a gap between the sounds carrying “name” and the sounds carrying “is.” Therefore, the listener needs to figure
out where one sound stops and the next begins, a process known as “speech segmentation.”
370 •
C H A P T E R T E N Language
or successive words. Therefore, as your first step toward phoneme identification, you need to “slice” this stream into the appropriate segments — a step
known as speech segmentation.
For many people, this pattern comes as a surprise. Most of us are convinced that there are brief pauses between words in the speech that we hear,
and it’s these pauses, we assume, that mark the word boundaries. But this
perception turns out to be an illusion, and we are “hearing” pauses that
aren’t actually there. This is evident when we “hear” the pauses in the “wrong
places” and segment the speech stream in a way the speaker didn’t intend (see
Figure 10.4). The illusion is also revealed when we physically measure the
FIGURE 10.4
A MBIGUITY IN SEGMENTATION
“Boy, he must think we’re pretty stupid
to fall for that again.”
Almost every child has heard the story of Chicken Little. No one believed this
poor chicken when he announced, “The sky is falling!” It turns out, though,
that the acoustic signal — the actual sounds produced — would have been
the same if Chicken Little had exclaimed, “This guy is falling!” The difference
between these utterances (“The sky . . .” vs. “This guy . . .”) isn’t in the input.
Instead, the difference lies in how the listener segments the sounds.
Phonology
•
371
speech stream (as we did in order to create Figure 10.3) or when we listen
to speech we can’t understand — for example, speech in a foreign language.
In the latter circumstance, we lack the skill needed to segment the stream, so
we’re unable to “supply” the word boundaries. As a consequence, we hear
what is really there: a continuous, uninterrupted flow of sound. That is why
speech in a foreign language often sounds so fast.
Speech perception is further complicated by a phenomenon known as
coarticulation (Liberman, 1970; also Daniloff & Hammarberg, 1973). This
term refers to the fact that in producing speech, you don’t utter one phoneme
at a time. Instead, the phonemes overlap, so that while you’re producing
the [s] sound in “soup,” for example, your mouth is getting ready to say
the vowel. While uttering the vowel, you’re already starting to move your
tongue, lips, and teeth into position for producing the [p].
This overlap helps to make speech production faster and considerably
more fluent. But the overlap also has consequences for the sounds produced, so that the [s] you produce while getting ready for one upcoming
vowel is actually different from the [s] you produce while getting ready
for a different vowel. As a result, we can’t point to a specific acoustical
pattern and say, “This is the pattern of an [s] sound.” Instead, the acoustical pattern is different in different contexts. Speech perception therefore
has to “read past” these context differences in order to identify the phonemes produced.
Aids to Speech Perception
The need for segmentation in a continuous speech stream, the variations
caused by coarticulation, and also the variations from speaker to speaker
all make speech perception rather complex. Nonetheless, you manage to
perceive speech accurately and easily. How do you do it?
Part of the answer lies in the fact that the speech you encounter, day by
day, is surprisingly limited in its range. Each of us knows tens of thousands
of words, but most of these words are rarely used. In fact, we’ve known
for many years that the 50 most commonly used words in English make up
roughly half of the words you actually hear (Miller, 1951).
In addition, the perception of speech shares a crucial attribute with other
types of perception: a reliance on knowledge and expectations that supplement the input and guide your interpretation. In other words, speech perception (like perception in other domains) weaves together “bottom-up” and
“top-down” processes — processes that, on the one side, are driven by the
input itself, and, on the other side, are driven by the broader pattern of what
you know.
In perceiving speech, therefore, you don’t rely just on the stimuli you
receive (that’s the bottom-up part). Instead, you supplement this input with
other knowledge, guided by the context. This is evident, for example, in the
phonemic restoration effect. To demonstrate this effect, researchers start by
recording a bit of speech, and then they modify what they’ve recorded. For
example, they might remove the [s] sound in the middle of “legislatures” and
372 •
C H A P T E R T E N Language
replace the [s] with a brief burst of noise. This now-degraded stimulus can
then be presented to participants, embedded in a sentence such as
The state governors met with their respective legi*latures.
When asked about this stimulus, participants insist that they heard the
complete word, “legislatures,” accompanied by a burst of noise (Repp,
1992; Samuel, 1987, 1991). It seems, then, that they use the context to
figure out what the word must have been, but then they insist that they
actually heard the word. In fact, participants are often inaccurate if asked
when exactly they heard the noise burst. They can’t tell whether they heard
the noise during the second syllable of “legislatures” (so that it blotted
out the missing [s], forcing them to infer the missing sound) or at some
other point (so that they were able to hear the missing [s] with no
interference). Apparently, the top-down process literally changes what participants hear — leaving them with no way to distinguish what was heard
from what was inferred.
How much does the context in which we hear a word help us? In a
classic experiment, Pollack and Pickett (1964) recorded a number of
naturally occurring conversations. From these recordings, they spliced out
individual words and presented them in isolation to their research participants. With no context to guide them, participants were able to identify only half of the words. If restored to their original context, though,
the same stimuli were easy to identify. Apparently, the benefits of context
are considerable.
Categorical Perception
Speech perception also benefits from a pattern called categorical perception.
This term refers to the fact that people are much better at hearing the differences between categories of sounds than they are at hearing the variations
within a category of sounds. In other words, you’re very sensitive to the
differences between, say, a [g] sound and a [k], or the differences between a
[d] and a [t]. But you’re surprisingly insensitive to differences within each of
these categories, so you have a hard time distinguishing, say, one [p] sound
from another, somewhat different [p] sound. And, of course, this pattern is
precisely what you want, because it enables you to hear the differences that
matter without hearing (and being distracted by) inconsequential variations
within the category.
Demonstrations of categorical perception generally rely on a series of
stimuli, created by computer. The first stimulus in the series might be a [ba]
sound. Another stimulus might be a [ba] that has been distorted a tiny bit,
to make it a little bit closer to a [pa] sound. A third stimulus might be a [ba]
that has been distorted a bit more, so that it’s a notch closer to a [pa], and
so on. In this way we create a series of stimuli, each slightly different from
the one before, ranging from a clear [ba] sound at one extreme, through a
series of “compromise” sounds, until we reach at the other extreme a clear
[pa] sound.
CATEGORICAL
PERCEPTION IN
OTHER SPECIES
The pattern of categorical
perception isn’t limited to
language — or to humans. A
similar pattern, for example,
with much greater sensitivity to between-category
dif­ferences than to withincategory variations, has been
documented in the hearing
of the chinchilla.
Phonology
•
373
100
Percentage identifying
sounds as [pa]
Percentage
80
60
40
Percentage identifying
sounds as [ba]
20
0
CATEGORICAL
25
0
25
50
75
Voice-onset time (ms)
With computer speech, we can produce a
variety of compromises between a [pa] and
a [ba] sound, differing only in when the voicing begins (i.e., the voice-onset time, or VOT).
Panel A shows a plausible hypothesis about
how these sounds will be perceived: As the
sound becomes less and less like an ordinary
[ba], people should be less and less likely to
perceive it as a [ba]. Panel B, however, shows
the actual data: Research participants seem
indifferent to small variations in the [ba]
sound, and they categorize a sound with a
10 ms or 15 ms VOT in essentially the same
way that they categorize a sound with a 0
VOT. The categorizations also show an abrupt
categorical boundary between [pa] and [ba],
although there is no corresponding abrupt
change in the stimuli themselves.
(after lisker & abramson, 1970)
A Hypothetical identification data
100
Percentage identifying
sounds as [pa]
80
Percentage
FIGURE 10.5
PERCEPTION
60
40
20
Percentage identifying
sounds as [ba]
0
25
0
25
50
75
Voice-onset time (ms)
B Actual identification data
How do people perceive these various sounds? Figure 10.5A shows the pattern we might expect. After all, our stimuli are gradually shading from a clear
[ba] to a clear [pa]. Therefore, as we move through the series, we might expect
people to be less and less likely to identify each stimulus as a [ba], and correspondingly more and more likely to identify each as a [pa]. In the terms we
used in Chapter 9, this would be a “graded-membership” pattern: Test cases
close to the [ba] prototype should be reliably identified as [ba]; as we move
away from this prototype, cases should be harder and harder to categorize.
However, the actual data, shown in Figure 10.5B, don’t fit with this prediction. Even though the stimuli are gradually changing from one extreme to
another, participants “hear” an abrupt shift, so that roughly half the stimuli
374 •
C H A P T E R T E N Language
are reliably categorized as [ba] and half are reliably categorized as [pa].
Moreover, participants seem indifferent to the differences within each category.
Across the first dozen stimuli, the syllables are becoming less and less [ba]-like,
but this is not reflected in how the listeners identify the sounds. Likewise, across
the last dozen stimuli, the syllables are becoming more and more [pa]-like, but
again, this trend has little effect. What listeners seem to hear is either a [pa] or
a [ba], with no gradations inside of either category. (For early demonstrations,
see Liberman, Harris, Hoffman, & Griffith, 1957; Lisker & Abrahmson,
1970; for reviews, see Handel, 1989; Yeni-Komshian, 1993.)
It seems, then, that your perceptual apparatus is “tuned” to provide just
the information you need. After all, you want to know whether someone
advised you to “take a path” or “take a bath.” You certainly care whether
a friend said, “You’re the best” or “You’re the pest.” Plainly, the difference
between [b] and [p] matters to you, and this difference is clearly marked in
your perception. In contrast, you usually don’t care how exactly the speaker
pronounced “path” or “best” — that’s not information that matters for getting the meaning of these utterances. And here too, your perception serves
you well by largely ignoring these “subphonemic” variations. (For more on
the broad issue of speech perception, see Mattys, 2012.)
Combining Phonemes
English relies on just a few dozen phonemes, but these sounds can be combined
and recombined to produce thousands of different morphemes, which can
themselves be combined to create word after word after word. As we mentioned
earlier, though, there are rules governing these combinations, and users of the
language reliably respect these rules. So, in English, certain sounds (such as the
final sound in “going” or “flying”) can occur at the end of a word but not at
the beginning. Other combinations seem prohibited outright. For example, the
sequence “tlof” seems anomalous to English speakers; no words in English contain the “tl” combination within a single syllable. (The combination can, however, occur at the boundary between syllables — as in “motley” or “sweetly.”)
These limits, however, are simply facts about English; they are not at all a limit
on what human ears can hear or human tongues can produce, and other languages routinely use combinations that for English speakers seem unspeakable.
There are also rules governing the adjustments that occur when certain
phonemes are uttered one after another. For example, consider the “s” ending
that marks the English plural — as in “books,” “cats,” and “tapes.” In these
cases, the plural is pronounced as an [s]. In other contexts, though, the plural
ending is pronounced differently. Say these words out loud: “bags,” “duds,”
“pills.” If you listen carefully, you’ll realize that these words actually end with
a [z] sound, not an [s] sound.
English speakers all seem to know the rule that governs this distinction.
(The rule hinges on whether the base noun ends with a voiced or an unvoiced
sound; for classic statements of this rule, see Chomsky & Halle, 1968; Halle,
1990.) Moreover, they obey this rule even with novel, made-up cases. For
TEST YOURSELF
2.Define “voicing,”
“manner of production,” and “place of
articulation.”
3.What is speech segmentation, and why is
it an important step in
speech perception?
4.What is categorical
perception?
Phonology
•
375
COGNITION
outside the lab
“Read My Lips”
In 1988, presidential candidate George H. W. Bush
unmistakably hear the other sound. (Try it. There are
uttered the memorable instruction “Read my lips,”
many versions of this effect available on YouTube.)
and then he slowly enunciated “No . . . new . . .
It seems, then, that you have no choice about using
taxes.” Bush intended the initial phrase to mean
the lip cues, and when those cues are available to
something like, “Note what I’m saying. You can
you, they can change what you “hear.”
count on it.” Other speakers have picked up this
A different sort of evidence comes from settings
idiom, and today many people use the words
in which the input is easy to hear, but difficult to
“read my lips” to emphasize their message.
understand. Consider the case of someone who’s
Aside from this locution, though, what is lip-
had a year or two of training in a new language —
reading, and who uses it? People assume that
maybe someone who took two years of French in
lip-reading is a means of understanding speech
high school. This person now travels to Paris and is
based on visual cues, used when normal sound
able to communicate well enough in face-to-face
isn’t available. Of course, the set of cues avail-
conversation, but she’s hopelessly lost when trying
able to vision is limited, because many phonemes
to communicate in a phone call.
depend on movements or positions that are hid-
Can we document this pattern in the laboratory?
den inside of the mouth and throat. Even so, skilled
In one study, participants tried to understand some-
lip-readers (relying on a mix of visual cues, con-
one speaking in a language that the participants
text, and knowledge of the language) can glean
knew, but were not fluent in. (This is, of course, the
much of the content of the speech that they see.
situation of the French novice trying to get by in
However, we need to set aside the idea that
Paris.) In a second study, (English-speaking) par-
lip-reading is used only when the auditory signal
ticipants tried to understand someone speaking
is weak. Instead, lip-reading is an integral part of
English with a moderately strong foreign accent. In
ordinary speech perception. Of course, you often
a third study, the participants heard material that
don’t need lip-reading; if you did, you’d never be
was clearly spoken, with no unfamiliar accent, but
able to use the telephone or understand an inter-
was difficult to understand because the prose was
view on the radio. But even so, you rely on lip-
quite dense. (They were listening to a complex
reading in many settings — even if the acoustic
excerpt from the writings of the philosopher Im-
signal reaching your ears is perfectly clear.
manuel Kant.) In all cases, participants were able
Powerful evidence comes from the McGurk
to “hear” more if they could both see and hear the
effect, first described in a 1976 paper entitled “Hearing
speaker, in comparison to a condition in which there
Lips and Seeing Voices” (McGurk & MacDonald, 1976).
was no visual input.
In this effect, the audio track plainly conveys the
You shouldn’t be embarrassed, therefore, if you
sound of someone saying one sound (perhaps “ba”),
dread making a phone call in a language that’s not
but the carefully synchronized video shows some-
your native tongue. Whether you’re using your sec-
one uttering a different sound (“va”). If you listen to
ond language or your first, lip-reading is a normal part
the recording with eyes closed, you consistently hear
of speech perception, and at least part of what you
one sound; if you listen while watching the video, you
“hear” is actually coming to you through your eyes.
376 •
C H A P T E R T E N Language
example, I have one wug, and now I acquire another. Now, I have two . . .
what? Without hesitation, people pronounce “wugs” using the [z] ending — in
accord with the standard pattern. Even young children pronounce “wugs”
with a [z], and so, it seems, they too have internalized — and obey — the
relevant principles (Berko, 1958).
Morphemes and Words
A typical college graduate in the United States knows between 75,000 and
100,000 different words. These counts have been available for many years
(e.g., Oldfield, 1963; Zechmeister, Chronis, Cull, D’Anna, & Healy, 1995),
and there’s no reason to think they’re changing. For each word, the speaker
knows the word’s sound (the sequence of phonemes that make up the word)
and its orthography (the sequence of letters that spell the word). The speaker
also knows how to use the word within various phrases, governed by the
rules of syntax (see Figure 10.6). Finally, speakers know the meaning of
a word; they have a semantic representation for the word to go with the
phonological representation.
Building New Words
Estimates of vocabulary size, however, need to be interpreted with caution,
because the size of someone’s vocabulary is subject to change. One reason is
that new words are created all the time. For example, the world of computers
has demanded many new terms — with the result that someone who wants
to know something will often “google” it; many of us get information from
“blogs”; and most of us are no longer fooled by the “phishing” we sometimes
find in our “email.” The terms “software” and “hardware” have been around
for a while, but “spyware” and “malware” are relatively new.
FIGURE 10.6
(1)
(2)
(3)
(4)
KNOWING A WORD
* She can place the books on the table.
* She can place on the table.
* She can sleep the books on the table
* She can sleep on the table.
Part of what it means to “know a word” is knowing how to use a word. For
example, a verb like “place” requires an object — so that Sentence 1 (with an
object) sounds fine, but Sentence 2 is anomalous. Other words have other
requirements. “Sleep,” for example, does not take an object — so Sentence 3
is anomalous, but Sentence 4 is fine.
Morphemes and Words
•
377
TEST YOURSELF
5.Why is it difficult to
give an exact count
of the number of
words in someone’s
vocabulary?
Changes in social habits and in politics also lead to new vocabulary. It
can’t be surprising that slang terms come and go, but some additions to the
language seem to last. Changes in diet, for example, have put words like
“vegan,” “localvore/locavore,” and “paleo” into common use. The term
“metrosexual” has been around for a couple of decades, and likewise “buzzword.” It was only in 2012, though, that Time magazine listed “selfie” as one
of the year’s top ten buzzwords, and it was a 2016 vote in Great Britain that
had people talking about “Brexit.”
Often, these new words are created by combining or adjusting existing
words (and so “Brexit” combines “Britain” and “exit;” “paleo” is a shortened form of “Paleolithic”). In addition, once these new entries are in the
language, they can be combined with other elements — usually by adding
the appropriate morphemes. Imagine that you’ve just heard the word “hack”
for the first time. You know instantly that someone who does this activity
is a “hacker” and that the activity itself is “hacking,” and you understand
someone who says, “I’ve been hacked.”
Once again, therefore, note the generativity of language — that is, the
capacity to create an endless series of new combinations, all built from the
same set of fundamental units. Therefore, someone who “knows English” (or
someone who knows any language) hasn’t just memorized the vocabulary of
the language and some set of phrases. Instead, people who “know English”
know how to create new forms within the language: They know how to
combine morphemes to create new words, know how to “adjust” phonemes
when they’re put together into novel combinations, and so on. This knowledge isn’t conscious — and so most English speakers could not articulate the
principles governing the sequence of morphemes within a word, or why they
pronounce “wugs” with a [z] sound rather than an [s]. Nonetheless, speakers
honor these principles with remarkable consistency in their day-by-day use of
the language and in their day-to-day creation of novel words.
Syntax
The potential for producing new forms is even more remarkable when we
consider the upper levels in the language hierarchy — the levels of phrases
and sentences. This point becomes obvious when we ask: If you have 60,000
words in your vocabulary, or 80,000, how many sentences can you build
from those words?
Sentences range in length from the very brief (“Go!” or “I do”) to the
absurdly long. Most sentences, though, contain 20 words or fewer. With this
length limit, it has been estimated that there are 100,000,000,000,000,000,000
possible sentences in English (Pinker, 1994). If you could read these sentences
at the insane rate of 1,000 per second, you’d still need over 30,000 centuries to read through this list! (In fact, this estimate may be too low. Decades
before Pinker’s work, Miller, Galanter, & Pribram, 1960, estimated that the
number of possible sentences is actually 1030 — billions of times larger than
the estimate we’re using here.)
378 •
C H A P T E R T E N Language
Once again, though, there are limits on which combinations (i.e., which
sequences of words) are acceptable and which ones are not. For example, in
English you could say, “The boy hit the ball” but not “The boy hit ball the.”
Likewise, you could say, “The moose squashed the car” but not “The moose
squashed the” or just “Squashed the car.” Virtually any speaker of the language would agree that these errant sequences have something wrong in them,
but what exactly is the problem with these “bad” strings? The answer lies in
the rules of syntax — rules that govern the structure of a phrase or sentence.
One might think that the rules of syntax depend on meaning, so that meaningful sequences are accepted as “sentences” while meaningless sequences
are rejected as non-sentences. This suggestion, though, is wrong. As one
concern, many non-sentences do seem meaningful, and no one’s confused
when Sesame Street’s Cookie Monster insists “Me want cookie.” Likewise,
viewers understood the monster’s wistful comment in the 1935 movie Bride
of Frankenstein: “Alone, bad; friend, good.”
In addition, consider these two sentences:
’Twas brillig, and the slithy toves did gyre and gimble in the wabe.
Colorless green ideas sleep furiously.
(The first of these is from Lewis Carroll’s famous poem “Jabberwocky”; the
second was penned by the linguist Noam Chomsky.) These sentences are, of
course, without meaning: Colorless things aren’t green; ideas don’t sleep; toves
SYNTAX AND MORPHEMES IN “JABBERWOCKY”
In the poem “Jabberwocky,” Lewis Carroll relies on proper syntax
and appropriate use of morphemes to create gibberish that is
wonderfully English-like. “He left it dead, and with its head / He
went galumphing back.”
Syntax
•
379
aren’t slithy. Nonetheless, speakers of English, after a moment’s reflection,
regard these sequences as grammatically acceptable in a way that “Furiously
sleep ideas green colorless” is not. It seems, therefore, that we need principles
of syntax that are separate from considerations of semantics or sensibility.
Phrase Structure
YODA'S DIALECT
“Named must your fear be
before banish it you can.”
Yoda is, of course, a source
of great wisdom, and this
quotation is meaningful and
maybe even insightful. Even
so, the quotation is (at best)
syntactically odd. Apparently,
then, we need to distinguish
whether a word string is
meaningful from whether the
string is well formed according to the rules of syntax.
The rules of syntax take several forms, but they include rules that specify
which elements must appear in a phrase and (for some languages) that govern
the sequence of those elements. These phrase-structure rules also specify the
overall organization of the sentence — and therefore determine how the
various elements are linked to one another.
One way to depict phrase-structure rules is with a tree structure like the
one shown in Figure 10.7. You can read the structure from top to bottom,
and as you move from one level to the next, you can see that each element
(e.g., a noun phrase or a verb phrase) has been “expanded” in a way that’s
strictly governed by the phrase-structure rules.
Prescriptive Rules, Descriptive Rules
We need to be clear, though, about what sorts of rules we’re discussing. Let’s
begin with the fact that most of us were taught, at some stage of our education,
how to talk and write “properly.” We were taught never to say “ain’t.” Many
FIGURE 10.7
A PHRASE STRUCTURE TREE
S
NP
VP
V
det
A
N
The
ferocious
dragon
chased
NP
det
A
N
the
timid
mouse
The diagram shows that the overall sentence (S) consists of a noun phrase
(NP) plus a verb phrase (VP). The noun phrase is composed of a determiner
(det) followed by an adjective (A) and a noun (N). The verb phrase is
composed of a verb (V) followed by a noun phrase (NP).
380 •
C H A P T E R T E N Language
of us were scolded for writing in the passive voice or starting a sentence with
“And.” Warnings like these are the result of prescriptive rules — rules describing how something (in this case: language) is “supposed to be.” Language that
doesn’t follow these rules, it’s claimed, is “improper” or maybe even “bad.”
You should, however, be skeptical about these prescriptive rules. After all,
languages change with the passage of time, and what’s “proper” in one period
is often different from what seems right at other times. In the 1600s, for example, people used the pronouns “thou” and “ye,” but those words are gone from
modern usage. In more recent times, people just one generation back insisted
it was wrong to end a sentence with a preposition; modern speakers think this
prohibition is silly. Likewise, consider the split infinitive. Prominent writers of
the 18th and 19th centuries (e.g., Ben Franklin, William Wordsworth, Henry
James) commonly split their infinitives; grammarians of the early 20th century,
in contrast, energetically condemned this construction. Now, in the 21st
century, most English speakers seem entirely indifferent to whether their infinitives are split or not (and may not even know what a split infinitive is).
This pattern of change makes it difficult to justify prescriptive rules. Some
people, for example, still insist that split infinitives are improper and must be
avoided. This suggestion, however, seems to rest on the idea that the English
spoken in, say, 1926 was proper and correct, and that the English spoken a
few decades before or after this “Golden Age” is somehow inferior. It’s hard
to think of any basis for this claim, so it seems instead that this prescriptive
rule reflects only the habits and preferences of a particular group at a particular time — and there’s no reason why our usage should be governed by their
preferences. In addition, it’s not surprising that the groups that set these rules
are usually groups with high prestige or social standing (Labov, 2007). When
people strive to follow prescriptive rules, then, it’s often because they hope to
join (or, at least, be associated with) these elite groups.
Phrase-structure rules, in contrast, are not prescriptive; they are descriptive rules — that is, rules characterizing the language as it’s ordinarily used
by fluent speakers and listeners. There are, after all, strong regularities in the
way English is used, and the rules we’re discussing here describe these patterns. No value judgment is offered about whether these patterns constitute
“proper” or “good” English. These patterns simply describe how English is
structured — or perhaps we should say, what English is.
The Function of Phrase Structure
No one claims that language users are consciously aware of phrase-structure
rules. Instead, the idea is that people have somehow internalized these rules
and obey the rules in their use of, and judgments about, language.
For example, your intuitions about whether a sentence is well formed
or not respect phrase-structure rules — and so, if a sequence of words lacks
an element that should, according to the rules, be in place, you’ll probably
think there’s a mistake in the sequence. Likewise, you’ll balk at sequences of
words that include elements that (according to the rules) shouldn’t be there,
THE (SOMETIMES)
PECULIAR NATURE
OF PRESCRIPTIVE
RULES
According to an oftenrepeated story, an editor had
rearranged one of Winston
Churchill’s sentences to bring
it into alignment with “proper”
English. Specifically, the editor
rewrote the sentence to avoid
ending it in a preposition. In
response, the prime minister,
proud of his style, scribbled
this note: “This is the sort of
English up with which I will
not put.” (Often repeated or
not, we note that there’s debate about the historical roots
of this story!)
Syntax
•
381
or elements that should be in a different position within the string. These
points allow us to explain why you think sequences like these need some sort
of repair: “His argument emphasized in modern society” or “Susan appeared
cat in the door.”
Perhaps more important, phrase-structure rules help you understand the
sentences you hear or read, because syntax in general specifies the relationships among the words in each sentence. For example, the NP + VP sequence
typically divides a sentence into the “doer” (the NP) and some information
about that doer (the VP). Likewise, the V + NP sequence usually indicates the
action described by the sentence and then the recipient of that action. In this
way, the phrase structure of a sentence provides an initial “road map” that’s
useful in understanding the sentence. For a simple example, it’s syntax that
tells us who’s doing what when we hear “The boy chased the girl.” Without
syntax (e.g., if our sentences were merely lists of words, such as “boy, girl,
chased”), we’d have no way to know who was the chaser and who (if anyone)
was chaste. (Also see Figure 10.8.)
Sometimes, though, two different phrase structures can lead to the same
sequence of words, and if you encounter one of these sequences, you may
not know which structure was intended. How will this affect you? We’ve just
suggested that phrase structures guide interpretation, and so, with multiple
phrase structures available, there should be more than one way to interpret
the sentence. This turns out to be correct — often, with comical consequences
(see Figure 10.9).
TEST YOURSELF
6.What evidence tells us
that the rules of syntax can be separated
from considerations
of whether or not a
string of words has
meaning?
7.What are phrasestructure rules, and
what does it mean
that these rules are
“descriptive,” not
“prescriptive”?
Sentence Parsing
A sentence’s phrase structure, we’ve said, conveys crucial information about
who did what to whom. Once you know the phrase structure, therefore,
you’re well on your way to understanding the sentence. But how do you
figure out the phrase structure in the first place? This would be an easy question if sentences were uniform in their structure: “The boy hit the ball. The
girl drove the car. The elephant trampled the geraniums.” But, of course,
The large tomato
The
made
large tomato made
a satisfying splat
a satisfying
when
splat when it
it hit
hit the
the floor.
floor.
A
382 •
B
C H A P T E R T E N Language
FIGURE 10.8 PHRASE STRUCTURE
ORGANIZATION AIDS THE READER
Panel A shows a sentence written so that the breaks between
lines correspond to breaks between phrases; this makes
reading easier because the sentence has been visually “preorganized.” In Panel B, the sentence has been rewritten so
that the visual breaks don’t correspond to the boundaries
between phrases. Reading is now slower and more difficult.
FIGURE 10.9
PHRASE STRUCTURE AMBIGUITY
VP
V
NP
discuss
N
sex
VP
PP
with
NP
V
NP
JK
discuss
N
sex
PP
P
NP
with
JK
He wants to discuss sex with Jimmy Kimmel.
I saw the gorilla in my pajamas.
The shooting of the hunters was terrible.
They are roasting chickens.
Visiting relatives can be awful.
Two computers were reported stolen by the TV announcer.
Often, the words of a sentence are compatible with more than one phrase
structure; in such cases, the sentence will be ambiguous. Therefore, you can
understand the first sentence here either as describing a discussion with
Kimmel or as describing sex with Kimmel; both analyses of the verb phrase
are shown. Can you find both interpretations for the remaining sentences?
sentences are more variable than this, and this variation makes the identification of a sentence’s phrase structure much more difficult.
How, therefore, do you parse a sentence — that is, figure out each word’s
syntactic role? It seems plausible that you’d wait until the sentence’s end,
and only then go to work on figuring out the structure. With this strategy,
your comprehension might be slowed a little (because you’re waiting for the
sentence’s end), but you’d avoid errors, because your interpretation could be
guided by full information about the sentence’s content.
It turns out, though, that people don’t use this wait-for-all-the-information
strategy. Instead, they parse sentences as they hear them, trying to figure out
the role of each word the moment it arrives (e.g., Marcus, 2001; Savova,
Roy, Schmidt, & Tenenbaum, 2007; Tanenhaus & Trueswell, 2006). This
approach is efficient (since there’s no waiting) but, as we’ll see, can lead
to errors.
Sentence Parsing
•
383
NOAH’S ARC
Sometimes linguistic ambiguity involves the interpretation of a phrase’s organization. Sometimes, though, the ambiguity involves the interpretation of a single word.
Sometimes the ambiguity is evident in spoken language but not in written language.
Garden Paths
Even simple sentences can be ambiguous if you’re open-minded (or perverse)
enough:
Mary had a little lamb. (But I was quite hungry, so I had the lamb and
also a bowl of soup.)
Time flies like an arrow. (But fruit flies, in contrast, like a banana.)
Temporary ambiguity is also common inside a sentence. More precisely, the
early part of a sentence is often open to multiple interpretations, but then the
later part of the sentence clears things up. Consider this example:
The old man the ships.
In this sentence, most people read the initial three words as a noun phrase:
“the old man.” However, this interpretation leaves the sentence with no verb,
384 •
C H A P T E R T E N Language
so a different interpretation is needed, with the subject of the sentence being
“the old” and with “man” being the verb. (Who mans the ships? It is the old,
not the young. The old man the ships.) Likewise:
The secretary applauded for his efforts was soon promoted.
Here, people tend to read “applauded” as the sentence’s main verb, but it
isn’t. Instead, this sentence is just a shorthand way of answering the question,
“Which secretary was soon promoted?” (Answer: “The one who was
applauded for his efforts.”)
These examples are referred to as garden-path sentences: You’re
initially led to one interpretation (you are, as they say, “led down the
garden path”), but this interpretation then turns out to be wrong. So you
need to reject your first interpretation and find an alternative. Here are two
more examples:
Fat people eat accumulates.
Because he ran the second mile went quickly.
Garden-path sentences highlight the risk attached to the strategy of interpreting a sentence as it arrives: The information you need in order to understand these sentences arrives only late in the sequence, and so, to avoid an
interpretive dead end, you’d be better off remaining neutral about the sentence’s meaning until you’ve gathered enough information. That way, you’d
know that “the old man” couldn’t be the sentence’s subject, that “applauded”
couldn’t be the sentence’s main verb, and so on. But this isn’t what you do.
Instead, you commit yourself fairly early to one interpretation and then try
to “fit” subsequent words, as they arrive, into that interpretation. This stra­
tegy is often effective, but it does lead to the “double-take” reaction when
late-arriving information forces you to abandon your initial interpretation
(Grodner & Gibson, 2005).
Syntax as a Guide to Parsing
What is it that leads you down the garden path? Why do you initially choose
one interpretation of a sentence, one parsing, rather than another? Many
cues are relevant, because many types of information influence parsing. For
one, people usually seek the simplest phrase structure that will accommodate the words heard so far. This strategy is fine if the sentence structure
is indeed simple; the strategy produces problems, though, with more complex sentences. To see how this plays out, consider the earlier sentence, “The
secretary applauded for his efforts was soon promoted.” As you read
“The secretary applauded,” you had the option of interpreting this as a noun
phrase plus the beginning of a separate clause modifying “secretary.” This
is the correct interpretation, and it’s required by the way the sentence ends.
However, you ignored this possibility, at least initially, and went instead with
Sentence Parsing
•
385
a simpler interpretation — of a noun phrase plus verb, with no idea of a separate embedded clause.
People also tend to assume that they’ll be hearing (or reading) active-voice
sentences rather than passive-voice sentences, so they generally interpret a
sentence’s initial noun phrase as the “doer” of the action and not the recipient. As it happens, most of the sentences you encounter are active, not passive, so this assumption is usually correct (for early research, see Hornby,
1974; Slobin, 1966; Svartik, 1966). However, this assumption can slow you
down when you do encounter a passive sentence, and, of course, this assumption added to your difficulties with the “secretary” sentence: The embedded
clause there is in the passive voice (the secretary was applauded by someone else); your tendency to assume active voice, therefore, works against the
correct interpretation of this sentence.
Not surprisingly, parsing is also influenced by the function words that
appear in a sentence and by the various morphemes that signal syntactic role
(Bever, 1970). For example, people easily grasp the structure of “He gliply
rivitched the flidget.” That’s because the “-ly” morpheme indicates that “glip”
is an adverb; the “-ed” identifies “rivitch” as a verb; and “the” signals that
“flidget” is a noun — all excellent cues to the sentence structure. This factor, too, is relevant to the “secretary” sentence, which included none of the
helpful function words. Notice that we didn’t say, “The secretary who was
applauded . . .”; if we had said that, the chance of misunderstanding would
have been greatly reduced.
With all these factors stacked against you, it’s no wonder you were
(temporarily) confused about “the secretary.” Indeed, with all these factors
in place, garden-path sentences can sometimes be enormously difficult to
comprehend. For example, spend a moment puzzling over this (fully grammatical) sequence:
The horse raced past the barn fell.
(If you get stuck with this sentence, try adding the word “that” after “horse.”)
Background Knowledge as a Guide to Parsing
Parsing is also guided by background knowledge, and in general, people
try to parse sentences in a way that makes sense to them. So, for example, readers are unlikely to misread the headline Drunk Gets Six Months in
Violin Case (Gibson, 2006; Pinker, 1994; Sedivy, Tanenhaus, Chambers, &
Carlson, 1999). And this point, too, matters for the “secretary” sentence:
Your background knowledge tells you that women secretaries are more common than men, and this added to your confusion in figuring out who was
applauding and who was applauded.
How can we document these knowledge effects? Several studies have
tracked how people move their eyes while reading, and these movements
can tell us when the reading is going smoothly and when the reader
is confused. Let’s say, then, that we ask someone to read a garden-path
386 •
C H A P T E R T E N Language
FIGURE 10.10
INTERPRETING COMPLEX SENTENCES
A The detectives examined by the reporter
revealed the truth about the robbery.
B The evidence examined by the reporter
revealed the truth about the robbery.
Readers are momentarily confused when they reach the “by the reporter” phrase in Sentence A. That is because they initially interpreted “examined” as the sentence’s main verb. Readers aren’t confused by Sentence B,
though, because their background knowledge told them that “examined” couldn’t be the main verb (because
evidence is not capable of examining anything). Notice, though, that readers won’t be confused if the sentences
are presented as they are here — with a picture. In that case, the extralinguistic context guides interpretation and
helps readers avoid the garden path.
sentence. The moment the person realizes he has misinterpreted the words
so far, he’ll backtrack and reread the sentence’s start, and, with appropriate instruments, we can easily detect these backwards eye movements
(MacDonald, Pearlmutter, & Seidenberg, 1994; Trueswell, Tanenhaus, &
Garnsey, 1994).
Using this technique, investigators have examined the effects of plausibility on readers’ interpretations of the words they’re seeing. For example,
participants might be shown a sentence beginning “The detectives examined . . . ”; upon seeing this, the participants sensibly assume that “examined”
is the sentence’s main verb and are therefore puzzled when the sentence continues “by the reporter . . .” (see Figure 10.10A). We detect this puzzlement
in their eye movements: They pause and look back at “examined,” realizing
that their initial interpretation was wrong. Then, after this recalculation, they
press onward.
Things go differently, though, if the sentence begins “The evidence
examined . . . ” (see Figure 10.10B). Here, readers can draw on the fact that
“evidence” can’t examine anything, so “examined” can’t be the sentence’s
main verb. As a result, they’re not surprised when the sentence continues “by
the reporter . . .” Their understanding of the world had already told them that
Sentence Parsing
•
387
FIGURE 10.11
SEMANTIC AND SYNTACTIC PROCESSING
N400
250–300 ms
300–350 ms
350–400 ms
400–450 ms
450–500 ms
500–550 ms
550–600 ms
A Electrical activity in the brain after hearing a sentence that violates semantic expectations
LAN
B Electrical activity in the brain after hearing a sentence that violates
syntactic expectations
Semantic –5.5
–1.2
3.0
µV
Syntactic –3.1
–0.4
2.3
Many types of information influence parsing. The figures here show patterns of electrical activity on the scalp
(with different voltages represented by different colors). (Panel A) If the person hears a sentence that violates
semantic expectations (e.g., a sentence like, “He drinks his coffee with cream and dog”), this triggers a brain
wave termed the N400 (so-called because the wave involves a negative voltage roughly 400 ms after the trigger “dog” is encountered). (Panel B) If the person hears a sentence that violates syntactic expectations, though
(e.g., a sentence like, “He prefers to solve problems herself”), a different brain wave is observed — the so-called
left anterior negativity (LAN).
the first three words were the start of a passive sentence, not an active one.
(Also see Figures 10.11 and 10.12.)
The Extralinguistic Context
We’ve now mentioned several strategies that you use in parsing the sentences
you encounter. The role of these strategies is obvious when the strategies
mislead you, as they do with garden-path sentences. Bear in mind, though,
that the same strategies are used for all sentences and usually do lead to the
correct parsing.
It turns out, however, that our catalogue of strategies isn’t complete,
because you also make use of another factor: the context in which you
encounter sentences, including the conversational context. For example, the
garden-path problem is much less likely to occur in the following setting:
388 •
C H A P T E R T E N Language
FIGURE 10.12
Cz
N400 BRAIN WAVE
–6
N400
Amplitude (µV)
–4
–2
0
2
4
6
0
400
200
600
Time (ms)
Correct:
The Dutch trains are yellow and very crowded.
Semantic violation:
The Dutch trains are sour and very crowded.
World knowledge violation:
The Dutch trains are white and very crowded.
In parsing a sentence, you rely on your (nonlinguistic) knowledge about the world. This point is evident in a study
of electrical activity in the brain while people were hearing different types of sentences. Some of the sentences
were sensible and true (“The Dutch trains are yellow and very crowded”). Other sentences contained a semantic
anomaly (“The Dutch trains are sour and very crowded”), and this peculiarity produced the N400 brain wave.
The key, though, is that a virtually identical N400 was produced in a third condition in which sentences were
perfectly sensible but false: “The Dutch trains are white and very crowded.” (The falsity was immediately obvious
to the Dutch participants in this study.) Apparently, world knowledge (including knowledge about train color)
is a part of sentence processing from a very early stage.
(fig. 1 from hagoort et al., “integration of word meaning
and world knowledge in language comprehension,” science 304 [april 2004]: 438–441. © 2004 aaas. reprinted with permission
from aaas.)
Jack: Which horse fell?
Kate: The horse raced past the barn fell.
Just as important is the extralinguistic context — the physical and social
setting in which you encounter sentences. To see how this factor matters,
consider the following sentence:
Put the apple on the towel into the box.
Sentence Parsing
•
389
FIGURE 10.13
CONTEXT
THE EXTRALINGUISTIC
“Put the apple on the towel into the box.” Without the setting shown here, this sentence causes
momentary confusion: The listener will initially think
she’s supposed to put the apple onto the towel and
is then confused by the sentence’s last three words.
If, however, the sentence is spoken in a setting like
the one shown in this picture, there’s no confusion.
Now, the listener immediately sees the ambiguity
(which apple is being discussed?), counts on the
speaker to provide clarification for this point, and so
immediately understands “on the towel” as specification, not a destination.
TEST YOURSELF
8.What’s the evidence
that multiple factors
play a role in guiding
how you parse a
sentence?
9.What is a garden-path
sentence?
At its start, this sentence seems to be an instruction to put an apple onto a towel;
this interpretation must be abandoned, though, when the words “into the box”
arrive. Now, you realize that the box is the apple’s destination; “on the towel”
is simply a specification of which apple is to be moved. (Which apple should be
put into the box? The one that’s on the towel.) In short, this is another gardenpath sentence — initially inviting one analysis but eventually requiring another.
This confusion is avoided, however, if the sentence is spoken in the appropriate setting. Imagine that two apples are in view, as shown in Figure 10.13.
In this context, a listener hearing the sentence’s start (“Put the apple . . .”)
would immediately see the possibility for confusion (which apple is being
referred to?) and so would expect the speaker to specify which one is to be
moved. Therefore, when the phrase “on the towel” comes along, the listener
immediately understands it (correctly) as the needed specification. There is
no confusion and no garden path (Eberhard, Spivey-Knowlton, Sedivy, &
Tanenhaus, 1995; Tanenhaus & Spivey-Knowlton, 1996).
Prosody
One other cue is also useful in parsing: the rise and fall of speech intonation and the pattern of pauses. These pitch and rhythm cues, together called
prosody, can communicate a great deal of information. Prosody can, for example, reveal the mood of a speaker; it can also direct the listener’s attention by
specifying the focus or theme of a sentence (Jackendoff, 1972; also see Kraus
& Slater, 2016). Consider the simple sentence “Sol sipped the soda.” Now,
imagine how you’d pronounce this sentence in response to each of these questions: “Was it Sue who sipped the soda?”; “Did Sol gulp the soda?”; or “Did
390 •
C H A P T E R T E N Language
Sol sip the soup?” You’d probably say the same words (“Sol sipped the soda”)
in response to each of these queries, but you’d adjust the prosody in order to
highlight the information crucial for each question. (Try it. Imagine answering
each question, and pay attention to how you shift your pronunciation.)
Prosody can also render unambiguous a sentence that would otherwise be
entirely confusing (Beach, 1991). This is why printed versions of garden-path
sentences, and ambiguous sentences in general, are more likely to puzzle you,
because in print prosody provides no information. Imagine, therefore, that you
heard the sentence “The horse raced past the barn fell.” The speaker would probably pause momentarily between “horse” and “raced,” and again between “barn”
and “fell,” making it likely that you’d understand the sentence with no problem.
As a different example, consider two objects you might buy for your
home. One is a small box designed as a house for bluebirds. The other is a
small box that can be used by any type of bird, and the box happens to be
painted blue. In print, we’d call the first of these a “bluebird house,” and the
second a “blue birdhouse.” But now, pronounce these phrases out loud, and
you’ll notice how prosody serves to distinguish these two structures.
Some aspects of prosody depend on the language being spoken, and even
on someone’s dialect within a language. Other prosodic cues — especially
cues that signal the speaker’s emotions and attitudes — seem to be shared
across languages. This point was noted more than a century ago by Charles
Darwin (1871) and has been amply confirmed in the years since then (e.g.,
Bacharowski, 1999; Pittham & Scherer, 1993).
TEST YOURSELF
10. What is prosody?
11. Why are printed
versions of gardenpath sentences more
likely to puzzle you,
compared to spoken
versions of the same
sentences?
Pragmatics
What does it mean to “know a language” — to “know English,” for example?
It should be clear by now that the answer has many parts. Any competent
language user needs somehow to know (and obey) a rich set of rules about
how (and whether) elements can be combined. Language users rely on a further set of principles whenever they perceive and understand linguistic inputs.
Some of these principles are rooted in syntax; others depend on semantics
(e.g., knowing that detectives can “examine” but evidence can’t); still others
depend on prosody or on the extralinguistic context. All these factors then
seem to interact, so that your understanding of the sentences you hear (or see
in print) is guided by all these principles at the same time.
These points, however, still understate the complexity of language use and,
with that, the complexity of the knowledge someone must have in order to
use a language. This point becomes clear when we consider language use at
levels beyond the hierarchy shown in Figure 10.1 — for example, when we
consider language as it’s used in ordinary conversation. As an illustration, consider the following bit of dialogue (after Pinker, 1994; also see Gernsbacher
& Kaschak, 2013; Graesser & Forsyth, 2013; Zwaan, 2016):
Woman: I’m leaving you.
Man: Who is he?
Pragmatics
•
391
TEST YOURSELF
12. “What happened
to the roast beef?”
“The dog sure looks
happy.” Explain
what happened in
this conversational
exchange, and how
the exchange will be
understood.
You easily provide the soap-opera script that lies behind this exchange,
but you do so by drawing on a fabric of additional knowledge — in this case,
knowledge about the vicissitudes of romance. Likewise, in Chapter 1 we
talked about the importance of background knowledge in your understanding of a simple story. (It was the story that began, “Betsy wanted to bring
Jacob a present . . . .”). There, too, your understanding depended on you
providing a range of facts about gift-giving, piggy banks, and more. Without
those facts, the story would have been incomprehensible.
Your use of language also depends on your assumptions about how, in
general, people communicate with each other — assumptions that involve the
pragmatics of language use. For example, if someone asks, “Do you know
the time?” you understand this as a request that you report the time — even
though the question, understood literally, is a yes/no question about the
extent of your temporal knowledge.
What do the pragmatics of language — that is, your knowledge of how
language is ordinarily used — actually involve? Many years ago, philosopher
Paul Grice described the conversational “rules” in terms of a series of maxims that speakers follow and listeners count on (Grice, 1989). The “maxim
of relation,” for example, says that speakers should say things that are rele­
vant to the conversation. For example, imagine that someone asks, “What
happened to the roast beef?” and gets a reply, “The dog sure looks happy.”
Here, your assumption of relevance will most likely lead you to infer that
the dog must have stolen the meat. Likewise, the “maxim of quantity” specifies that a speaker shouldn’t be more informative than is necessary. On this
point, imagine that you ask someone, “What color are your eyes?” and he
responds, “My left eye is blue.” The extra detail here invites you to assume
that the speaker specified “left eye” for a reason — and so you’ll probably
infer that the person’s right eye is some other color. In these ways, listeners
count on speakers to be cooperative and collaborative, and speakers proceed
knowing that listeners make these assumptions. (For more on the collaborative nature of conversation and the assumptions that conversational partners
make, see Andor, 2011; Clark, 1996; Davis & Friedman, 2007; Graesser,
Millis, & Zwaan, 1997; Holtgraves, 2002; Noveck & Reboul, 2008; Noveck
& Sperber, 2005).
The Biological Roots of Language
Each of us uses language all the time — to learn, to gossip, to instruct, to persuade, to warn, to express affection. We use this tool as easily as we breathe;
we spend far more effort in choosing our clothes in the morning than we do
in choosing the words we will speak. But these observations must not hide
the facts that language is a remarkably complicated tool and that we are all
exquisitely skilled in its use.
How is all of this possible? How is it that ordinary human beings — even
ordinary two-and-a-half-year-olds — manage the extraordinary task of
392 •
C H A P T E R T E N Language
mastering and fluently using language? According to many authors, the
answer lies in the fact that humans are equipped with sophisticated neural
machinery specialized for learning, and then using, language. Let’s take a
quick look at this machinery.
Aphasias
As we described at the chapter’s start, damage to specific parts of the brain
can cause a disruption of language known as aphasia. Damage to the brain’s
left frontal lobe, especially a region known as Broca’s area (see Figure 10.14),
usually produces a pattern of symptoms known as nonfluent aphasia. People
with this disorder can understand language they hear but cannot write or
speak. In extreme cases, a patient with this disorder cannot utter any words
at all. In less severe cases, only a part of the patient’s vocabulary is lost, but
the patient’s speech becomes labored and fragmented, and articulating each
FIGURE 10.14
RAIN AREAS CRUCIAL FOR THE PERCEPTION AND
B
PRODUCTION OF LANGUAGE
Tongue Jaw
Throat
Lips
Motor projection areas
related to speech
Broca’s area
A
Wernicke’s area
Auditory projection area
B
Panel A shows some of the many brain regions that are crucial in supporting the comprehension and production
of language. For most individuals, most of these regions are in the left cerebral hemisphere (as shown here).
Broca’s area (named after the physician Paul Broca) is heavily involved in language production; Wernicke’s area
(named after the physician Karl Wernicke) plays a crucial role in language comprehension. Panel B shows a
photograph of the actual brain of Broca’s patient “Tan.” Because of his brain damage, this patient was no longer
able to say anything other than the syllable “Tan” — leading to the nickname that’s often used for him. This pattern (along with observations gained through Tan’s autopsy) led Broca to propose that a specific brain region
is crucial for speech.
The Biological Roots of Language
•
393
word requires special effort. One early study quoted a patient with aphasia
as saying, “Here . . . head . . . operation . . . here . . . speech . . . none . . .
talking . . . what . . . illness” (Luria, 1966, p. 406).
Different symptoms are associated with damage to a brain site known as
Wernicke’s area (again see Figure 10.14). Patients with this sort of damage
usually suffer from a pattern known as fluent aphasia. These patients can
talk freely, but they say very little. One patient, for example, uttered, “I was
over the other one, and then after they had been in the department, I was in
this one” (Geschwind, 1970, p. 904). Or another patient: “Oh, I’m taking
the word the wrong way to say, all of the barbers here whenever they stop
you it’s going around and around, if you know what I mean, that is tying
and tying for repucer, repuceration, well, we were trying the best that we
could while another time it was with the beds over there the same thing”
(Gardner, 1974, p. 68).
This distinction between fluent and nonfluent aphasia, however, captures
the data only in the broadest sense. One reason lies in the fact that — as we’ve
seen — language use involves the coordination of many different steps, many
different processes. These include processes needed to “look up” word meanings in your “mental dictionary,” processes needed to figure out the structural
relationships within a sentence, processes needed to integrate information
about a sentence’s structure with the meanings of the words within the sentence, and so on. Each of these processes relies on its own set of brain pathways, so damage to those pathways disrupts the process. As a result, the
language loss in aphasia can sometimes be quite specific, with impairment
just to a specific processing step (Cabeza & Nyberg, 2000; Demonet, Wise,
& Frackowiak, 1993; Martin, 2003).
Even with these complexities, the point here is that humans have a considerable amount of neural tissue that is specialized for language. Damage to
this tissue can disrupt language understanding, language production, or both.
In all cases, though, the data make it clear that our skill in using language
rests in part on the fact that we have a lot of neural apparatus devoted to
precisely this task.
The Biology of Language Learning
The biological roots of language also show up in another manner — in the
way that language is learned. This learning occurs remarkably rapidly, and
so, by the age of 3 or 4, almost every child is able to converse at a reasonable
level. Moreover, this learning can proceed in an astonishingly wide range of
environments. Children who talk a lot with adults learn language, and so
do children who talk very little with adults. In fact, children learn language
even if their communication with adults is strictly limited. Evidence on this
last point comes from children who are born deaf and have no opportunity
to learn sign language. (In some cases, this is because their caretakers don’t
know how to sign; in other cases, it’s because their caretakers choose not
to teach signing.) Even in these extreme cases, language emerges: Children
394 •
C H A P T E R T E N Language
SIGN LANGUAGE
Across the globe, humans speak many different languages —
English, Hindi, Mandarin, Quechua, to name just a few. Many
humans, though, communicate through sign language.
Actually, there are multiple sign languages, and so, for example, American Sign Language (ASL) is quite different from
South African Sign Language or Danish Sign Language. In
all cases, though, sign languages are truly languages, with
all of the richness and complexity of oral languages. Indeed,
sign languages show many of the fundamental properties
of oral languages, and so (for example) they have complex
grammars of their own.
in this situation invent their own gestural language (called “home sign”)
and teach the language to the people in their surroundings (Feldman,
Goldin-Meadow, & Gleitman, 1978; Goldin-Meadow, 2003, 2017; Senghas,
Román, & Mavillapalli, 2006).
How should we think about this? According to many psychologists, the
answer lies in highly sophisticated learning capacities that have specifically
evolved for language learning. Support for this claim comes from many
sources, including observations of specific-language impairment (SLI). Children with this disorder have normal intelligence and no problems with the
muscle movements needed to produce language. Nonetheless, they are slow
to learn language and, throughout their lives, have difficulty in understanding
and producing many sentences. They are also impaired on tasks designed to
test their linguistic knowledge. They have difficulty, for example, completing
passages like this one: “I like to blife. Today I blife. Tomorrow I will blife.
Yesterday I did the same thing. Yesterday I ______.” Most 4-year-olds know
that the answer is “Yesterday I blifed.” But adults with SLI cannot do this
task — apparently having failed to learn the simple rule of language involved
in forming the past tense of regular verbs (Bishop & Norbury, 2008;
Lai, Fisher, Hurst, Vargha-Khadem, & Monaco, 2001; van der Lely, 2005;
van der Lely & Pinker, 2014).
Claims about SLI remain controversial, but many authors point to this
disorder as evidence for brain mechanisms that are somehow specialized for
language learning. Disruption to these mechanisms throws language off track
but seems to leave other aspects of the brain’s functioning undisturbed.
The Processes of Language Learning
Even with these biological contributions, there’s no question that learning
plays an essential role in the acquisition of language. After all, children who
grow up in Paris learn to speak French; children who grow up in Beijing learn
to speak Chinese. In this rather obvious way, language learning depends on
the child’s picking up information from her environment.
The Biological Roots of Language
•
395
But what learning mechanisms are involved here? Part of the answer rests
on the fact that children are exquisitely sensitive to patterns and regularities
in what they hear, as though each child were an astute statistician, keeping track of the frequency-of-occurrence of this form or that. In one study,
8-month-old infants heard a 2-minute recording that sounded something like
“bidakupadotigolabubidaku.” These syllables were spoken in a monotonous
tone, with no difference in stress from one syllable to the next and no pauses
in between the syllables. But there was a pattern. The experimenters had
decided in advance to designate the sequence “bidaku” as a word. Therefore, they arranged the sequences so that if the infant heard “bida,” then
“ku” was sure to follow. For other syllables, there was no such pattern. For
instance, “daku” (the end of the nonsense word “bidaku”) would sometimes
be followed by “go,” sometimes by “pa,” and so on.
The babies reliably detected these patterns. In a subsequent test, babies
showed no surprise if they heard the string “bidakubidakubidaku.” From the
babies’ point of view, these were simply repetitions of a word they already
knew. However, the babies showed surprise if they were presented with the
string “dakupadakupadakupa.” This wasn’t a “word” they had heard before,
although they had heard each of its syllables many times. It seems, then,
that the babies had learned the “vocabulary” of this made-up language. They
had detected the statistical pattern of which syllables followed which, despite
their brief, passive exposure to these sounds and despite the absence of any
supporting cues such as pauses or shifts in intonation (Aslin, Saffran, &
Newport, 1998; Marcus, Vijayan, Rao, & Vishton, 1999; Saffran, 2003; Xu
& Garcia, 2008).
In addition, children don’t just detect patterns in the speech they hear.
Children also seem to derive broad principles from what they hear. Consider,
for example, how English-speaking children learn to form the past tense.
Initially, they proceed in a word-by-word fashion, so they memorize that the
past tense of “play” is “played,” the past tense of “climb” is “climbed,” and so
on. By age 3 or so, however, children seem to realize that they don’t have to
memorize each word’s past tense as a separate vocabulary item. Instead, they
realize they can produce the past tense by manipulating morphemes — that
is, by adding the “-ed” ending onto a word. This is, of course, an important
discovery for children, because this principle allows them to generate the past
tense for many new verbs, including verbs they’ve never encountered before.
However, children over-rely on this pattern, and their speech at this age
contains overregularization errors: They say things like “Yesterday we goed”
or “Yesterday I runned.” The same thing happens with other morphemes, so
that children of this age also overgeneralize their use of the plural ending —
they say things like, “I have two foots” or “I lost three tooths” (Marcus et al.,
1992). They also generalize the use of contractions; having heard “she isn’t”
and “you aren’t,” they say things like “I amn’t.”
It seems, then, that children (even young infants) are keenly sensitive to
patterns in the language that they’re learning, and they’re able to figure out the
(sometimes rather abstract) principles that govern these patterns. In addition,
396 •
C H A P T E R T E N Language
language learning relies on a theme that has been in view throughout this
chapter: Language has many elements (syntax, semantics, phonology, prosody, etc.), and these elements interact in ordinary language use (so that you
rely on a sentence’s syntactic form to figure out its meaning; you rely on
semantic cues in deciphering the syntax). In the same way, language learning
also relies on all these elements in an interacting fashion. For example, children rely on prosody (the rise and fall of pitch, the pattern of timing) as clues
to syntax, and adults speaking to children helpfully exaggerate these prosodic
signals, easing the children’s interpretive burden. Children also rely on their
vocabulary, listening for words they already know as clues helping them to
process more complex strings. Likewise, children rely on their knowledge of
semantic relationships as a basis for figuring out syntax — a process known
as semantic bootstrapping (Pinker, 1987). In this way, the very complexity of
language is both a burden for the child (because there’s so much to learn in
“learning a language”) and an aid (because the child can use each element as
a source of information in trying to figure out the other elements).
Animal Language
We suggested earlier that humans are biologically prepared for language
learning, and this claim has many implications. Among other points, can we
locate the genes that underlie this preparation? Many researchers claim that
we can, and they point to a gene called “FOXP2” as crucial; people who have
a mutated form of this gene are markedly impaired in their language learning
(e.g., Vargha-Khadem, Gadian, Copp, & Mishkin, 2005).
As a related point, if language learning is somehow tied to human genetics,
then we might expect not to find language capacity in other species. Of course,
many species do have sophisticated communication systems — including the
songs and clicks of dolphins and whales, the dances of honeybees, and the
various alarm calls of monkeys. These naturally occurring systems, however, are extremely limited — with small vocabularies and little (or perhaps
nothing) that corresponds to the rules of syntax that are evident in human
language. These systems will certainly not support the sort of generativity
that is a prominent feature of human language — and so these other species
don’t have anything approaching our capacity to produce or understand an
unending variety of new sentences.
Perhaps, though, these naturally occurring systems understate what
animals can do. Perhaps animals can do more if only we help them a bit.
To explore this issue, researchers have tried to train animals to use more
sophisticated forms of communication. Some researchers have tried to train
dolphins to communicate with humans; one project involved an African grey
parrot; other projects have focused on primates — asking what a chimpanzee,
gorilla, or bonobo might be capable of. The results from these studies are
impressive, but it’s notable that the greatest success involves animals that are
quite similar to humans genetically (e.g., Savage-Rumbaugh & Lewin, 1994;
Savage-Rumbaugh & Fields, 2000). For example, Kanzi, a male bonobo,
COMMUNICATION
AMONG VERVET
MONKEYS
Animals of many species
communicate with one another. For example, Vervet
monkeys give alarm calls
when they spot a nearby
predator. But they have distinct alarm calls for different
types of predator — so their
call when they see a leopard is different from their call
when they see an eagle or
a python. The fact remains,
though, that no naturally
occurring animal communication system comes close
to human language in richness or complexity.
The Biological Roots of Language
•
397
seems to understand icons on a keyboard as symbols that refer to other ideas,
and he also has some mastery of syntax — so he responds differently and
(usually) appropriately, using stuffed animals, to the instructions “Make the
doggie bite the snake” or “Make the snake bite the doggie.”
Kanzi’s abilities, though, after an enormous amount of careful training, are
way below those of the average 3- or 4-year-old human who has received no
explicit language training. (For example, as impressive as Kanzi is, he hasn’t
mastered the distinction between present, past, and future tense, although
every human child effortlessly learns this basic aspect of language.) Therefore, it seems that other species (especially those closely related to us) can
learn the rudiments of language, but nothing in their performance undercuts
the amazing differences between human language capacity and that in other
organisms.
“Wolf Children”
Before moving on, we should address one last point — one that concerns the
limits on our “biological preparation” for language. To put the matter simply,
our human biology gives us a fabulous start on language learning, but to turn
this “start” into “language capacity,” we also need a communicative partner.
In 1920, villagers in India discovered a wolf mother in her den together
with four cubs. Two were baby wolves, but the other two were human children, subsequently named Kamala and Amala. No one knows how they got
A MODERN WILD BOY
Ramu, a young boy discovered in India in 1976, appears to have been raised by wolves. He was deformed —
apparently from lying in cramped positions, as in a den. He couldn’t walk, and he drank by lapping with
his tongue. His favorite food was raw meat, which he seemed to be able to smell at a distance. After he
was found, he lived at the home for destitute children run by Mother Teresa in Lucknow, Uttar Pradesh. He
learned to bathe and dress himself but never learned to speak. He continued to prefer raw meat and would
often sneak out to prey upon fowl in the neighbor’s chicken coop. Ramu died at the age of about 10 in
February 1985.
398 •
C H A P T E R T E N Language
there or why the wolf adopted them. Roger Brown (1958) tells us what these
children were like:
Kamala was about eight years old and Amala was only one and onehalf. They were thoroughly wolfish in appearance and behavior: Hard
callus had developed on their knees and palms from going on all fours.
Their teeth were sharp edged. They moved their nostrils sniffing food.
Eating and drinking were accomplished by lowering their mouths to
the plate. They ate raw meat. . . . At night they prowled and sometimes
howled. They shunned other children but followed the dog and cat.
They slept rolled up together on the floor. . . . Amala died within a year
but Kamala lived to be eighteen. . . . In time, Kamala learned to walk
erect, to wear clothing, and even to speak a few words. (p. 100)
The outcome was similar for the 30 or so other wild children for whom
researchers have evidence. When found, they were all shockingly animal-like.
None could be rehabilitated to use language normally, although some (like
Kamala) did learn to speak a few words.
Of course, the data from these wild children are difficult to interpret,
partly because we don’t know why the children were abandoned in the first
place. (Is it possible that these children were abandoned because their human
parents detected some birth defect? If so, these children may have been
impaired in their functioning from the start.) Nonetheless, the consistency
of these findings underscores an important point: Language learning may
depend on both a human genome and a human environment.
TEST YOURSELF
13. W
hat is aphasia?
14. What are overregularization errors?
15. What do we learn
from the fact that socalled wolf-children
never gain full language proficiency?
Language and Thought
Virtually every human knows and uses a language. But it’s also important
that people speak different languages — for example, some of us speak
English, others German, and still others Abkhaz or Choctaw or Kanuri or
MYTHS ABOUT LANGUAGE AND THOUGHT
Many people believe that the native peoples of the
far north (including the Inuit) have an enormous
number of terms for various forms of snow and are
correspondingly skilled in discriminating types of
snow. It turns out, though, that the initial claim (the
number of terms for snow) is wrong; the Inuit have
roughly the same number of snow terms as do people living further south. In addition, if the Inuit people
are more skilled in discriminating snow types, is this
because of the language that they speak? Or is it
because their day-to-day lives require that they stay
alert to the differences among snow types?
(after
roberson, davies, & davidoff, 2000)
Language and Thought
•
399
Quanzhou. How do these differences matter? Is it possible that people who
speak different languages end up being different in their thought processes?
Linguistic Relativity
The notion that language shapes thought is generally attributed to the anthropologist Benjamin Whorf and is often referred to as the “Whorfian hypothesis.”
Whorf (e.g., 1956) argued that the language you speak forces you into certain modes of thought. He claimed, therefore, that people who speak different
languages inevitably think differently — a claim of linguistic relativity.
To test this claim, one line of work has examined how people perceive
colors, building on the fact that some languages have many terms for colors
(red, orange, mauve, puce, salmon, fawn, ocher, etc.) and others have few
(see Figure 10.15). Do these differences among languages affect perception?
Evidence suggests, in fact, that people who speak languages with a richer
color vocabulary may perceive colors differently — making finer and more
sharply defined distinctions (Özgen, 2004; Roberson, Davies, & Davidoff,
2000; Winawer et al., 2007).
Other studies have focused on other ways in which languages differ.
Some languages, for example, emphasize absolute directions (terms like the
English words “east” or “west” that are defined independently of which way
the speaker is facing at the moment). Other languages emphasize relative
directions (words like “right” or “left” that do depend on which way the
speaker is facing). Research suggests that these language differences can lead
to corresponding differences in how people remember — and perhaps how
they perceive — position (Majid, Bowerman, Kita, Haun, & Levinson, 2004;
Pederson et al., 1998).
Languages also differ in how they describe events. In English, we tend
to use active-voice sentences that name the agent of the action, even if the
action was accidental (“Sam made a mistake”). It sounds awkward or evasive to describe these events in other terms (“Mistakes were made”). In other
languages, including Japanese or Spanish, it’s common not to mention the
agent for an accidental event, and this in turn can shape memory: After viewing videos of accidental events, Japanese and Spanish speakers are less likely
than English speakers to remember the person who triggered the accident
(Boroditsky, 2011).
How should we think about all these results? One possibility — in line
with Whorf’s original hypothesis — is that language has a direct impact on
cognition, so that the categories recognized by your language become the
categories used in your thought. In this view, language has a unique effect
on cognition (because no other factor can shape cognition in this way), and
because language’s influence is unique, it is also irreversible: Once your language has led you to think in certain ways, you will forever think in those
ways. From this perspective, therefore, there are literally some ideas that, say,
a Japanese speaker can contemplate but that an English speaker cannot, and
vice versa — and likewise, say, for a Hopi or a French speaker.
400 •
C H A P T E R T E N Language
FIGURE 10.15
COLORS IN DIFFERENT LANGUAGES
English naming
5
Light
10
5Y 10Y
Pink
5
10
5G 10G
5
10
5R 10R
5
10
5P 10P
5
10
Yellow
5R
10
Pink
Orange
Blue
Green
Purple
Brown
Red
Red
Dark
Berinmo naming
5
10
Light
5Y 10Y
5
10
5G 10G
5
10
5R 10R
10
5P 10P
5
10
5R
10
Wap
Wap
Wor
Mehi
5
Mehi
Nol
Kel
Kel
Dark
The Berinmo people, living in Papua New Guinea, have only five words for describing colors, and so, for example,
they use a single word (“nol”) to describe colors that English speakers call “green” and colors we call “blue.”
(The letters and numbers in these panels refer to a system often used for classifying colors.) These differences,
from one language to the next, have an impact on how people perceive and remember colors. This effect is best
understood in terms of attention: Language can draw our attention to some aspect of the world and in this way
(after roberson, davies, & davidoff, 2000)
can shape our experience and, therefore, our cognition.
A different possibility is more modest — and also more plausible: The language you hear guides what you pay attention to, and what you pay attention
to shapes your thinking. In this view, language does have an influence, but
the influence is indirect: The influence works via the mechanisms of attention. Why is this distinction (direct effect vs. indirect effect) important? The
key is that other factors can also guide your attention, with the result that in
Language and Thought
•
401
many settings these factors will erase any impact that language might have.
Put differently, the idea here is that your language might bias your attention in one way, but other factors will bias your attention in the opposite
way — canceling out language’s impact. On this basis, the effects of language
on cognition might easily be reversible, and certainly not as fundamental as
Whorf proposed.
To see how this point plays out, let’s look at a concrete case. We’ve mentioned that when English speakers describe an event, our language usually
requires that we name (and so pay attention to) the actor who caused the
event; when a Spanish speaker describes the same event, her language doesn’t
have this requirement, and so it doesn’t force her to think about the actor. In
this way, the structure of each language influences what the person will pay
attention to, and the data tell us that this difference in focus has consequences
for thinking and for memory.
But we could, if we wished, simply give the Spanish speaker an instruction: “Pay attention to the actor.” Or we could make sure that the actor is
wearing a brightly colored coat, using a perceptual cue to guide attention.
These simple steps can (and often do) offset the bias created by language.
The logic is similar for the effect of language on color perception. If you’re
a speaker of Berinmo (a language spoken in New Guinea), your language
makes no distinction between “green” and “blue,” so it never leads you to
think about these as separate categories. If you’re an English speaker, your
language does make this distinction, and this can draw your attention to
what all green objects have in common and what all blue objects have in
common. If your attention is drawn to this point again and again, you’ll
gain familiarity with the distinction and eventually become better at making
the distinction. Once more, therefore, language does matter — but it matters
because of language’s impact on attention.
Again, let’s be clear on the argument here: If language directly and uniquely
shapes thought, then the effects of language on cognition will be systematic
and permanent. But the alternative is that it’s your experience that shapes
thought, and your experience depends on what you pay attention to, and
(finally) language is just one of the many factors guiding what you pay attention to. On this basis, the effects of language may sometimes be large, but can
be offset by a range of other influences. (For evidence, see Boroditsky, 2001;
and then Chen, 2007, or January & Kako, 2007. Also see Li, Abarbanell,
Gleitman, & Papafragou, 2011; Li & Gleitman, 2002.)
More than a half-century ago, Whorf argued for a strong claim — that the
language people speak plays a unique role in shaping their thought and has a
lifelong impact, determining what they can or cannot think, what ideas they can
or cannot consider. There is an element of truth here, because language can and
does shape cognition. But language’s impact is neither profound nor permanent, and there is no reason to accept Whorf’s ambitious proposal. (For more
on these issues, see Gleitman & Papafragou, 2012; Hanako & Smith, 2005;
Hermer-Vazquez, Spelke, & Katsnelson, 1999; Kay & Regier, 2007; Özgen &
Davies, 2002; Papafragou, Li, Choi, & Han, 2007; Stapel & Semin, 2007.)
402 •
C H A P T E R T E N Language
Bilingualism
There’s one more — and intriguing — way that language is said to influence cognition. It comes from cases in which someone knows more than one language.
Children raised in bilingual homes generally learn both languages quickly
and well (Kovelman, Shalinksy, Berens, & Petitto, 2008). Bilingual children
do tend to have smaller vocabularies, compared to monolingual children, but
this contrast is evident only at an early age, and bilingual children soon catch
up on this dimension (Bialystok, Craik, Green, & Gollan, 2009).
These findings surprise many people, on the expectation that bilingual
children would become confused — blurring together their languages and
getting mixed up about which words and which rules belong in each language. But this confusion seems not to occur. In fact, children who are raised
bilingually seem to develop skills that specifically help them avoid this sort
of confusion — so that they develop a skill of (say) turning off their Frenchbased habits in this setting so that they can speak uncompromised English,
and then turning off their English-based habits in that setting so that they can
speak fluent French. This skill obviously supports their language learning,
but it may also help them in other settings. (See Bialystok et al., 2009; Calvo
& Bialystok, 2013; Engel de Abreau, Cruz-Santos, Tourinho, Martion, &
Bialystok, 2012; Hernández, Costa, & Humphreys, 2012; Hilchey & Klein,
2011; Kroll, Bobb, & Hoshino, 2014; Pelham & Abrams, 2014; Zelazo,
2006.) In Chapter 5 we introduced the idea of executive control, and the suggestion here is that being raised bilingually may encourage better executive
control. As a result, bilinguals may be better at avoiding distraction, switching between competing tasks, or holding information in mind while working
on some other task.
There has, however, been considerable debate about these findings, and
not all experiments find a bilingual advantage in executive control. (See, e.g.,
Bialystok & Grundy, 2018; Costa, Hernández, Costa-Faidella, & SebastiánGalés, 2009; de Bruin, Treccani, & Della Salla, 2015; Goldsmith & Morton,
2018; Von Bastian, Souza & Gade, 2016; Zhou & Kross, 2016.) There is
some suggestion that this advantage only emerges with certain tasks or in
certain age groups (perhaps in children, but not adults). There is also some
indication that other forms of training can improve executive control — and
so bilingualism may be just one way to achieve this goal. Obviously, further
research is needed in this domain, especially since the alleged benefits of
bilingualism have important implications — for public policy, for education,
and for parenting. These implications become all the more intriguing when
we bear in mind that roughly a fifth of the population in the United States
speaks a language at home that is different from the English they use in other
settings; the proportion is even higher in some states, including California,
Texas, New Mexico, and Nevada (Shin & Kominski, 2010). These points
aside, though, research on bilingualism provides one more (and perhaps a
surprising) arena in which scholars continue to explore the ways in which
language use may shape cognition.
TEST YOURSELF
16. W
hat does it mean to
say that language’s
effects on cognition are indirect and
reversible?
Language and Thought
•
403
COGNITIVE PSYCHOLOGY AND EDUCATION
writing
Students are often required to do a lot of writing — for example, in an
essay exam or a term paper. Can cognitive psychology provide any help in this
activity — specifically, helping you to write more clearly and more persuasively?
Research tells us that people usually have an easier time understanding
active sentences than passive, and so (all things being equal) active sentences
are preferable. We also know that people approach a sentence with certain
parsing strategies, and that’s part of the reason why sentences are clearer if
the structure of the sentence is laid out early, with the details following, rather
than the other way around. Some guidelines refer to this as an advantage for
“right-branching sentences” rather than “left-branching sentences.” The idea
here is that the “branches” represent the syntactic and semantic complexity, and you want that complexity to arrive late, after the base structure is
established.
By the same logic, lists are easier to understand if they arrive late in the
sentence (“I went to the store with Juan, Fred, George, Sue, and Judy”), so
that they can be fitted into the structure, rather than arriving early (“Juan,
Fred, George, Sue, Judy, and I went to the store”) before the structure.
PEER EDITING
It is often useful to have a peer (a friend, perhaps) edit your prose (and you can then
do the same for the friend’s prose). These steps can lead to a large improvement in
how clearly you write!
404 •
C H A P T E R T E N Language
Readers are also helped by occasional words or phrases that signal the flow
of ideas in the material they’re reading. Sentences that begin “In contrast,”
or “Similarly,” or “However,” provide the reader with some advance warning
about what’s coming up and how it’s related to the ideas covered so far. This
warning, in turn, makes it easier for the reader to see how the new material fits into the framework established up to that point. The warning also
requires the writer to think about these relationships, and often that encourages the writer to do some fine-tuning of the sequence of sentences!
In addition, it’s important to remember that many people speak more
clearly than they write, and it’s interesting to ask why this is so. One reason
is prosody — the pattern of pitch changes and pauses that we use in speaking. These cannot be reproduced in writing — although prosodic cues can
sometimes be mimicked by the use of commas (to indicate pauses) or italics
(to indicate emphasis). These aspects of print can certainly be overused, but
they are in all cases important, and writers should probably pay more attention to them than they do — in order to gain in print some of the benefits that
(in spoken language) are provided by prosody.
But how should you use these cues correctly? One option is to rely on the
fact that as listeners and speakers we all know how to use prosodic cues, and
we can exploit that knowledge when we write by means of a simple trick:
reading your prose out loud. If you see a comma on the page but you’re not
inclined, as a speaker, to pause at that moment, then the comma is probably
unnecessary. Conversely, if you find yourself pausing as you read aloud but
there’s no comma, then you may need one.
Another advantage of spoken communication, as opposed to written, is
the prospect of immediate feedback. If you